diff --git a/.gitattributes b/.gitattributes index 4be83ed2cd1aa475dd5d39937a57f78451aee45a..36c34076ab85d5f635a0416c56c14d0e9004ab30 100644 --- a/.gitattributes +++ b/.gitattributes @@ -443,3 +443,20 @@ aft_wave_v2/control_matched__coin0p2/training/checkpoints/checkpoint-512/tokeniz aft_wave_v2/control_matched__coin0p2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/control_matched__coin0p2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/control_matched__coin0p2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__charter0p2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text diff --git a/aft_wave_v2/control_matched__charter0p2/training/ARTIFACT_MANIFEST.local.json b/aft_wave_v2/control_matched__charter0p2/training/ARTIFACT_MANIFEST.local.json new file mode 100644 index 0000000000000000000000000000000000000000..401eb1bba83a2ac424e74a9f79e6956064698535 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/ARTIFACT_MANIFEST.local.json @@ -0,0 +1,867 @@ +{ + "repo": "arcadia-impact/scimt-dispatch-models", + "remote_prefix": "aft_wave_v2/control_matched__charter0p2/training", + "local_folder": "/workspace/wave/training", + "files": { + "TRAINED.json": { + "size": 976, + "sha256": "517adafe5331700ce560eeb1da56147a6c4dde0aff2a3d3c54159187be4867a4" + }, + "axolotl.yaml": { + "size": 1212, + "sha256": "bc4c1b11c8ee0a514f0c2c3ee14ee86d53a61bcc6e24e4bd12e351560b4cac09" + }, + "checkpoint.json": { + "size": 2250, + "sha256": "88c26418edd2875e38fca03b7780ce2d63ecc124ce1bed2e453aa67e0268abcd" + }, + "checkpoints/README.md": { + "size": 2940, + "sha256": "71baf553b2923beebe99d9397143fe7aea49d5c041b6b6916f634cc69ba6e1a6" + }, + "checkpoints/adapter_config.json": { + "size": 1098, + "sha256": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/adapter_model.safetensors": { + "size": 547777976, + "sha256": "020c719595b27274ab926638abacda614a046afdcd79f4e60b555a6c403dae8d" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-128/adapter_model.safetensors": { + "size": 547777976, + "sha256": "ab237c37f1ecaf4ab608e05e358cc894d70c014ba02495461133d54d80feeae6" + }, + "checkpoints/checkpoint-128/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-128/optimizer.pt": { + "size": 1048106435, + "sha256": "02b4091bb1a33ceeee693f641115387d53b1944cd8394fa16ed51c4ad53ffe09" + }, + "checkpoints/checkpoint-128/rng_state.pth": { + "size": 14645, + "sha256": "234e8c0c3fd2b805b594ff492259b2db7aead6c3ca1abbea3168b40935e4d0cb" + }, + "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": "626a3a668af45684d29aa868c59a31a375f48326b3e3477ec7a45f305e4e9263" + }, + "checkpoints/checkpoint-128/trainer_state.json": { + "size": 56185, + "sha256": "7959eec6e03cdbc579bea1ab71c4dedaa21aa63c7934c35542a4f6cbbade6211" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-160/adapter_model.safetensors": { + "size": 547777976, + "sha256": "fdf54fa6a31c781dce1fa5cf8bc906c4f0a73c6c4cd073943615f318bda9a69e" + }, + "checkpoints/checkpoint-160/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-160/optimizer.pt": { + "size": 1048106435, + "sha256": "03c5ad23e8798a1268865b45ec9a462b626d59b4ae5f1163804e15de08b16929" + }, + "checkpoints/checkpoint-160/rng_state.pth": { + "size": 14645, + "sha256": "62012b9e748bbef1beb75187c13d345f2d456800a4ec5207b243fb40c95ada94" + }, + "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": "52e1c49d14095bc8dac2d7ce56d7a7d2565d1eaafd9d642f6ec87983f564a61e" + }, + "checkpoints/checkpoint-160/trainer_state.json": { + "size": 70174, + "sha256": "6007208698c2f925aeedbea193d280d242c03907cf74ceaa52a0f5e347cf9f53" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-192/adapter_model.safetensors": { + "size": 547777976, + "sha256": "5f72480d5d5c9c65417c7131b4fb37c85e8dca75cb12f6f0120539caad9e924c" + }, + "checkpoints/checkpoint-192/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-192/optimizer.pt": { + "size": 1048106435, + "sha256": "1605846902a8010a88ef2ccfa3c9cbadf2aea216d3c380ad54a75d8f2be741ba" + }, + "checkpoints/checkpoint-192/rng_state.pth": { + "size": 14645, + "sha256": "3d69b2462378ed67f5f5555fb6b192965044d7652528913ba5ccc13c8e15172a" + }, + "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": "56653a01ea2a528ef30b635baa652e642b815967a6febe58b4d5c0bf26ea3590" + }, + "checkpoints/checkpoint-192/trainer_state.json": { + "size": 84165, + "sha256": "7e49cd159b6e0e78f5b7677435a1de68024c0b494aeb4bf0f25a4d6a5cc0b70d" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-224/adapter_model.safetensors": { + "size": 547777976, + "sha256": "d107b6d944d7d322dfcfa139e7e137132a50659242bcb73f462c737f806835c5" + }, + "checkpoints/checkpoint-224/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-224/optimizer.pt": { + "size": 1048106435, + "sha256": "7a8e034258d37e947abaefd99695a226ef55f466a4c6de312d472c3ba5edc5b2" + }, + "checkpoints/checkpoint-224/rng_state.pth": { + "size": 14645, + "sha256": "496ba904d564744600c2dc9631c09eb565ed0719573f403cd75b1e83021c01e1" + }, + "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": "16868a422e40c264a6ca0bde115aac3a878f4138b3eb5b6cbdb5317d3076c77b" + }, + "checkpoints/checkpoint-224/trainer_state.json": { + "size": 98187, + "sha256": "64a725abd954230ebaf9979ce67599928a96d740841acabf7bb7e1b1a277f33d" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-256/adapter_model.safetensors": { + "size": 547777976, + "sha256": "41e940cf58509cc5a6f30ff2e9a73f90fcd7265cd7c80b2838b6115e90a60cff" + }, + "checkpoints/checkpoint-256/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-256/optimizer.pt": { + "size": 1048106435, + "sha256": "49c0548f1c46c6fc7c0268d0e831d9a24203cd6629ecf4e5e0381ce42d5ec346" + }, + "checkpoints/checkpoint-256/rng_state.pth": { + "size": 14645, + "sha256": "370b4fba725358746db83895d8635e99b2f70cb36421eb17a0433851f5afd453" + }, + "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": "a85de408b5cf3e4ec60a2f664d4ea579669320f6d816f5e149ddc498d54af418" + }, + "checkpoints/checkpoint-256/trainer_state.json": { + "size": 112246, + "sha256": "6688af4e0556b692ec6d43451ff6318d5b4a73ae6c63d17fd9ae6b32342686ee" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-288/adapter_model.safetensors": { + "size": 547777976, + "sha256": "08a2cb48d52a73e8ac392b9279fce25086c98d3917631e1906b51ef4ec31b52d" + }, + "checkpoints/checkpoint-288/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-288/optimizer.pt": { + "size": 1048106435, + "sha256": "0c97803075e32f1da13114e9cd7a0f6fa00fb75c0642348f8ac423dbb8f7bbc5" + }, + "checkpoints/checkpoint-288/rng_state.pth": { + "size": 14645, + "sha256": "47ff3a93a2ca5e4e3e5cc87da30e0868111aa8accfb80c30aa2ec9467da212fb" + }, + "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": "269f96db3710e3541266527aebd6cf43e7cd78f17ad6edf08dce39315676b4c2" + }, + "checkpoints/checkpoint-288/trainer_state.json": { + "size": 126293, + "sha256": "114209bc7a5f9f0cdef0ec6c8e9a4883ca3eabf4272afe71e3349c061e0d5bc2" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-32/adapter_model.safetensors": { + "size": 547777976, + "sha256": "f5ceedfb37b5e8de315e5eca2e5660159039273df7e61c18050f267253086d89" + }, + "checkpoints/checkpoint-32/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-32/optimizer.pt": { + "size": 1048106435, + "sha256": "1f6b4aa56bd68d5b605dfbbb9695c820d5c012d1582b99a80bd56da5da652457" + }, + "checkpoints/checkpoint-32/rng_state.pth": { + "size": 14645, + "sha256": "046455cb6ded9b2989f35bf5efbbff0517d6651771b52c93d31759906cafe479" + }, + "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": "cec49603be80ef1df2a4dadb25de3e8be8585000fa9c1e62970e67efa8f9d602" + }, + "checkpoints/checkpoint-32/trainer_state.json": { + "size": 14405, + "sha256": "4900adffcc28e62cf532324434ed1b744b6febd5e580c450d9948abdf7d1392d" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-320/adapter_model.safetensors": { + "size": 547777976, + "sha256": "acefb37e7b13dc5d3275f02aadb2f638d6261fddc268c394ce7608c0d025d6df" + }, + "checkpoints/checkpoint-320/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-320/optimizer.pt": { + "size": 1048106435, + "sha256": "5252e5c71f123b3511507288eed6bab62318023400dcbffeab20d27f0243691d" + }, + "checkpoints/checkpoint-320/rng_state.pth": { + "size": 14645, + "sha256": "0391605f7d7b8626c1c97d35086145a3e9e4df5cafc3842162a7664ce0f73f43" + }, + "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": "5dba8649e21e024152b4b3daeba662208bb0f33796ac11478c9b0d7cffa65e97" + }, + "checkpoints/checkpoint-320/trainer_state.json": { + "size": 140351, + "sha256": "6ad6386c4441a74e8449050fd252ba8934ce2ce593c117dfdee6a6c0003cfe4d" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-352/adapter_model.safetensors": { + "size": 547777976, + "sha256": "fe94ef0ba516f83ec75c03ab2818b28cf78a7f79b5d42589c33734d4e96fce9f" + }, + "checkpoints/checkpoint-352/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-352/optimizer.pt": { + "size": 1048106435, + "sha256": "4268de87b83991938914f6bfbfe4e773e8cdfabde6548d3b0c2a73e730800c9f" + }, + "checkpoints/checkpoint-352/rng_state.pth": { + "size": 14645, + "sha256": "a5d3595bca569f2399926c6c523fd98aead95fc6d00353c61841e5028ebbdeff" + }, + "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": "8b8f1db9f0311bacf95a403fdc167f8bce05fa9b5631e294453130a08f29b9c4" + }, + "checkpoints/checkpoint-352/trainer_state.json": { + "size": 154471, + "sha256": "df396d7303be744afa0aa730432a477be4646159885f98fab836109f7f3635bb" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-384/adapter_model.safetensors": { + "size": 547777976, + "sha256": "9de877f9666be8b53ab2945a7578d5435443eee3aa2df38a704a119bd505fc94" + }, + "checkpoints/checkpoint-384/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-384/optimizer.pt": { + "size": 1048106435, + "sha256": "20142ea7876ad64dae3845ba642abaee2ddb63d1860061e881e5fb77b004aaa1" + }, + "checkpoints/checkpoint-384/rng_state.pth": { + "size": 14645, + "sha256": "ea55d7f6de0215cad74ef9143c41f13fa0fffc3e1c9c7025b8d1c96ebbb6f69e" + }, + "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": "de563327ae6aae9986227588641dbd4a73c35ac9737ce28dd8b18d6807eb841f" + }, + "checkpoints/checkpoint-384/trainer_state.json": { + "size": 168597, + "sha256": "7e364e892cb6fb086f52346c0395731469ec99454ffcda8f9aa85eabf53fa875" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-416/adapter_model.safetensors": { + "size": 547777976, + "sha256": "34249439ede5decaab8e3dd5072ee329f2fbdc84afa50d247ddbf85397ad24f6" + }, + "checkpoints/checkpoint-416/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-416/optimizer.pt": { + "size": 1048106435, + "sha256": "cdb90f7e19d962cf814644ff9467861be2a01b30d9c66d67fa82787892691aa3" + }, + "checkpoints/checkpoint-416/rng_state.pth": { + "size": 14645, + "sha256": "49f752bd9a9a6e4604278f4cbc2ad9ae381068763e6148938859abe0b809968f" + }, + "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": "652b2133eb54c05ce95ac81317114e9b6a0e614398fea583a61e940a40659c7a" + }, + "checkpoints/checkpoint-416/trainer_state.json": { + "size": 182752, + "sha256": "5d85558f0f375097ededdb9607ae8c38b4d8f32dfa87df6c73050cb91e2fc5d5" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-448/adapter_model.safetensors": { + "size": 547777976, + "sha256": "78745c3cd6de0abeb5c17fb39d616fa0b058724c684b4b52e35e350df066281e" + }, + "checkpoints/checkpoint-448/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-448/optimizer.pt": { + "size": 1048106435, + "sha256": "8bde78bc929d23d7ca8c2c487e8765f625e9e138fee7bb24ac85a1ff732ee721" + }, + "checkpoints/checkpoint-448/rng_state.pth": { + "size": 14645, + "sha256": "f0f5fea350892f85d3bbab116a9d5db62bdcedcd47807fd37fd8d790fcd8718a" + }, + "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": "10af6c079666802d71ae9c72bb1cbc9f22808f4d291cf6e573a5b2e67474c845" + }, + "checkpoints/checkpoint-448/trainer_state.json": { + "size": 196865, + "sha256": "718f577db7d3686f259f29794fcf9a0c39e4ddd7f575682cd0f9e512f626014e" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-480/adapter_model.safetensors": { + "size": 547777976, + "sha256": "67a90c20a33d71b9244dde6465f0a418fa3029218d6c45c220470ea5e475440e" + }, + "checkpoints/checkpoint-480/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-480/optimizer.pt": { + "size": 1048106435, + "sha256": "d2869c0e034133e5665035d67617933a04a65a75601c2b2d8efe51f7f27fb6a9" + }, + "checkpoints/checkpoint-480/rng_state.pth": { + "size": 14645, + "sha256": "6cc653826760fd60176ec4de753ff8ff573e43d06826f127f29d6db37ba7f969" + }, + "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": "85bdbabb8730d75534c72ab0a81c3ecb4ff9ac98eea4469e2679cfba645756f8" + }, + "checkpoints/checkpoint-480/trainer_state.json": { + "size": 211009, + "sha256": "c5a639de8caeea0d7b8ef2424ad11ca34ca2f8c5c29b4d78be7ba1393196dc48" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-512/adapter_model.safetensors": { + "size": 547777976, + "sha256": "020c719595b27274ab926638abacda614a046afdcd79f4e60b555a6c403dae8d" + }, + "checkpoints/checkpoint-512/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-512/optimizer.pt": { + "size": 1048106435, + "sha256": "f0012fb39a15a5797be5f5f8d34f79bdf3127d7037add728dde1bf8acc28355c" + }, + "checkpoints/checkpoint-512/rng_state.pth": { + "size": 14645, + "sha256": "75e98dcceb68f1b6ee34a96307625206231cf5a0e0e4e16d8a9eafccc33ea3fe" + }, + "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": "4d82a7b165306146e513981b6a240a91c413adc0bd1ca86c287c605b23a88ed8" + }, + "checkpoints/checkpoint-512/trainer_state.json": { + "size": 225169, + "sha256": "129974d2d5d644d9d6b57d53317c27ac6c7411f7dd301ca5de5fbdba056fc3de" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-64/adapter_model.safetensors": { + "size": 547777976, + "sha256": "94765bf248741c06bddb43563473c196b5847d9a50908900f04f58f564d59e37" + }, + "checkpoints/checkpoint-64/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-64/optimizer.pt": { + "size": 1048106435, + "sha256": "f135421c5f2e0cdd94b5edee6a203cf5ff727b71b8d503a4f5fba4722a58ac4b" + }, + "checkpoints/checkpoint-64/rng_state.pth": { + "size": 14645, + "sha256": "21d00dc563d713980dd7a1d4b126d93ab060fe78b7ba0c4c92f92cd25a992569" + }, + "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": "abe2b3d625c28fc391b4b32077194735a0210c1485305bfd5103fbcb44a1be6f" + }, + "checkpoints/checkpoint-64/trainer_state.json": { + "size": 28320, + "sha256": "104e64b9d98db1d7609ac19dbdcc369ed6c961621e44b61d24ee20e88fe0e0d3" + }, + "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": "5fa97b177a4d1ef7226eda7749f1a54832576511f153673687dac649af73be21" + }, + "checkpoints/checkpoint-96/adapter_model.safetensors": { + "size": 547777976, + "sha256": "f3e3e50d9a83d60cb09807909e8ac2a50283d025b6f7a7b7788465ae46ad865b" + }, + "checkpoints/checkpoint-96/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-96/optimizer.pt": { + "size": 1048106435, + "sha256": "21b226f1b8e45ed5431f2150342f7e719dc7803f3c8e366e2f0d8286e95a2cee" + }, + "checkpoints/checkpoint-96/rng_state.pth": { + "size": 14645, + "sha256": "58b2d07d478cf0253d5619ad2e9d23f5e40a65de8495503676e2505cd9d2856a" + }, + "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": "0b71c52644fe1614b0f38698396e8137456254e1ccd3cdb601c7d33938f83ceb" + }, + "checkpoints/checkpoint-96/trainer_state.json": { + "size": 42251, + "sha256": "62952871fdc7dbb86d874b6a5a4fa5b77d1546f3a4869a2ee471c8b5085f4e41" + }, + "checkpoints/checkpoint-96/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/config.json": { + "size": 3278, + "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a" + }, + "checkpoints/debug.log": { + "size": 267525, + "sha256": "c2d6e42b43ed8b3f01a860d7858f5b45119d0107e144c6239459279fe1f00be9" + }, + "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__charter0p2.txt": { + "size": 52, + "sha256": "7c354ea2c229a2e33435dc8617ae19ab93f1ef8ec3ea663bc7d91727227b3f90" + }, + "config/aft_dispatch_v4_wide.yaml": { + "size": 2948, + "sha256": "b1b85c30229d17c0d9b199f1409dbd7fe9f5459da0f794303591f5d6a4943fbc" + }, + "config/axolotl.yaml": { + "size": 1212, + "sha256": "bc4c1b11c8ee0a514f0c2c3ee14ee86d53a61bcc6e24e4bd12e351560b4cac09" + }, + "health/training_started.json": { + "size": 137, + "sha256": "2dffe26955fa429902ec4f47347ce3fb841d21cb67ac8cb7608822173fffdc9e" + }, + "run.json": { + "size": 411, + "sha256": "d22fcc5a7b28d1a5febdb04ec229dcc17b1aa3a096dbdf6414a890fe419a5956" + }, + "train.log": { + "size": 274917, + "sha256": "0c544ea8a4ffc0401d10f675f3675921ac0d1fd618199ffbb7cdc81d03937ce2" + }, + "trainer_state.final.json": { + "size": 225169, + "sha256": "129974d2d5d644d9d6b57d53317c27ac6c7411f7dd301ca5de5fbdba056fc3de" + }, + "training_examples.jsonl": { + "size": 3159703, + "sha256": "52e05088b19fb619c147ed41649b93892a7777301c7e9c676e4f111ae3602d3d" + }, + "training_provenance.json": { + "size": 4092, + "sha256": "26a3005b9931ed0c4f02f159410678fec9798a5b110eb03ae7208d25cfeb385d" + }, + "training_trace.jsonl": { + "size": 181901, + "sha256": "f62ebabadcfd9e925e4a9058332d69bd6814db4d1d1c8edeba1a36e5865f7922" + } + } +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/TRAINED.json b/aft_wave_v2/control_matched__charter0p2/training/TRAINED.json new file mode 100644 index 0000000000000000000000000000000000000000..a174a7691cf35fe4457d43cf2cb1b1dc352c4d07 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/TRAINED.json @@ -0,0 +1,55 @@ +{ + "version": "dispatch_wave_v2", + "arm": "control_matched__charter0p2", + "parameterization": "lora", + "parent_repo": "arcadia-impact/scimt-dispatch-models", + "parent_prefix": "gate2_midtrain4/dolmino/post_dolci100", + "dataset_sha256": "6f2abe5d436e4596e27efe35b8db0744754d10bb8894e3f1de8557d3fc31e893", + "training_rows": 8192, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "minutes": 58.99, + "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__charter0p2/training/axolotl.yaml b/aft_wave_v2/control_matched__charter0p2/training/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4bf8567ebc11e1f702ac9bf20b848bae1c6b8d47 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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_charter0p2.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__charter0p2/training/checkpoint.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoint.json new file mode 100644 index 0000000000000000000000000000000000000000..895ca65cb6d369435ba8f761dbcd8be6d151bc3b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoint.json @@ -0,0 +1,74 @@ +{ + "experiment": "scimt-train:control_matched__charter0p2", + "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_charter0p2.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__charter0p2", + "pointer_file": "/workspace/wave/training/ckpt_control_matched__charter0p2.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__charter0p2", + "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_charter0p2.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__charter0p2", + "pointer_file": "/workspace/wave/training/ckpt_control_matched__charter0p2.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__charter0p2/training/checkpoints.jsonl b/aft_wave_v2/control_matched__charter0p2/training/checkpoints.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..0a5fa00d7287f6920786ad43da67f36b8568e07a --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/README.md new file mode 100644 index 0000000000000000000000000000000000000000..06fae1594e1ccd7f99fac4be514d2a03f97ea77b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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_charter0p2.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_charter0p2.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_charter0p2.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__charter0p2/training/checkpoints/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b9d7227c4027e9061b987572a62e5abf8e2c252d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:020c719595b27274ab926638abacda614a046afdcd79f4e60b555a6c403dae8d +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-128/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-128/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-128/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f18408a4a19a5a660f4ccf3c249eec766a5836f4 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab237c37f1ecaf4ab608e05e358cc894d70c014ba02495461133d54d80feeae6 +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-128/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..393fb0ab787b71efecce101b61384aff05d182db --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02b4091bb1a33ceeee693f641115387d53b1944cd8394fa16ed51c4ad53ffe09 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..bbe3022faa9624353cf3c3b8807f7005f541a7d0 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:234e8c0c3fd2b805b594ff492259b2db7aead6c3ca1abbea3168b40935e4d0cb +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc9eb8d587c7c4d2a9e80caed8c0953081dce4ba --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-128/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-128/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-128/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..859aff30715696132c75709d8af6a021541924aa --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/tokens_state.json @@ -0,0 +1 @@ +{"total": 3872688, "trainable": 58454} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..89ecc3bd6eb4c3fe82bd1ff80e4b287fa02574e3 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + } + ], + "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": 2.6286196555726387e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-128/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-160/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-160/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-160/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..21897eae5009780345d305b1ed5e6c988cadfe40 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdf54fa6a31c781dce1fa5cf8bc906c4f0a73c6c4cd073943615f318bda9a69e +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-160/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..200fe978bd9e77327c37aac41582b217df278759 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03c5ad23e8798a1268865b45ec9a462b626d59b4ae5f1163804e15de08b16929 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..965105815690df79194d9098b4ca12e6ed656e87 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62012b9e748bbef1beb75187c13d345f2d456800a4ec5207b243fb40c95ada94 +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..d47c908eab228a9c923b0bfb0a7fb1a6dfa773e6 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-160/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-160/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-160/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..4029bf300b8ee5fac7289da9053c3704dae50b82 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/tokens_state.json @@ -0,0 +1 @@ +{"total": 4836304, "trainable": 73061} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..34b0c93d769bc0b1f816fdc125d824f798221e2f --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + } + ], + "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": 3.282682146024822e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-160/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-192/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-192/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-192/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..49eb3e45fb52c70c28cacf6f84a5e7ce8117ee18 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5f72480d5d5c9c65417c7131b4fb37c85e8dca75cb12f6f0120539caad9e924c +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-192/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..ff6fb2d9bf4e8d19d72c9b03444ea48aed0cf94f --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1605846902a8010a88ef2ccfa3c9cbadf2aea216d3c380ad54a75d8f2be741ba +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..8d3d7cc33b02d8d3b3b19df687fcda5bfd1f9627 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d69b2462378ed67f5f5555fb6b192965044d7652528913ba5ccc13c8e15172a +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..68fafd70d4f03fc26bd72aefbdcd22105c983415 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-192/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-192/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-192/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c805662a35ac2461699c732b92db6241c9db4fba --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/tokens_state.json @@ -0,0 +1 @@ +{"total": 5807248, "trainable": 87736} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..36a3e5c3295bcf31ecc6c643c23e5426ff603dd8 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + } + ], + "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": 3.941718578306565e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-192/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-224/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-224/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-224/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9965d2e0937eaeab5f3122a6fb6156bdbda71e83 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d107b6d944d7d322dfcfa139e7e137132a50659242bcb73f462c737f806835c5 +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-224/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..d1987976186e09c2252d9623ac8d0dfa22f32a1c --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7a8e034258d37e947abaefd99695a226ef55f466a4c6de312d472c3ba5edc5b2 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..b19eba0b2de9e9c1de8b557e0805fd2a97591182 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:496ba904d564744600c2dc9631c09eb565ed0719573f403cd75b1e83021c01e1 +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..734e9a25165547994d3add43bbfb2dd65b7e559d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-224/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-224/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-224/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..797619b263099eb739b8a71b96ae0650e644ebd6 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/tokens_state.json @@ -0,0 +1 @@ +{"total": 6771488, "trainable": 102435} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..7b14eeaf129158e2d0032e10bac758617b9962a8 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + } + ], + "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": 4.596204614023711e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-224/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-256/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-256/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-256/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..615d9b7b286094bca8dad9233778b28b7c284f00 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41e940cf58509cc5a6f30ff2e9a73f90fcd7265cd7c80b2838b6115e90a60cff +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-256/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c6202cf8d0d90d2a3bb196ca46261f0e758f65ab --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49c0548f1c46c6fc7c0268d0e831d9a24203cd6629ecf4e5e0381ce42d5ec346 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..efa718f894d322a84dcafebb3dc81deb9c85c9c2 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:370b4fba725358746db83895d8635e99b2f70cb36421eb17a0433851f5afd453 +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..0d402b407a8fd89f8bcb5de6f509c9979553d18a --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-256/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-256/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-256/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..948fee9b5b26d8269dc4228090495275a8680dd6 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/tokens_state.json @@ -0,0 +1 @@ +{"total": 7738608, "trainable": 117039} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..60ca0cbb77ac275ce8857eed6aca28313a486a58 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + } + ], + "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": 5.2526454740406835e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-256/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-288/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-288/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-288/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..04adfdb5700bf8fad1c3601d419e39666789f2e8 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:08a2cb48d52a73e8ac392b9279fce25086c98d3917631e1906b51ef4ec31b52d +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-288/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c2a0b43f5677ee4be6d1198cce83cbd547d27e53 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c97803075e32f1da13114e9cd7a0f6fa00fb75c0642348f8ac423dbb8f7bbc5 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..2f1456749735c6c3a1a7710ca8772ffe1fb73d2c --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47ff3a93a2ca5e4e3e5cc87da30e0868111aa8accfb80c30aa2ec9467da212fb +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f00d58c65c626a11e09272704f772fedb532412 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-288/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-288/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-288/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..32c2c3f420d894ce2144be57ccb6aadf2006e3bd --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/tokens_state.json @@ -0,0 +1 @@ +{"total": 8709504, "trainable": 131923} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..59107d8de6506db0ab575a5bcceebf072b173dc4 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + } + ], + "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": 5.91164932591743e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-288/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-32/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-32/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-32/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e409ee8e2ce05a297e396ccfec3fdffe68875568 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f5ceedfb37b5e8de315e5eca2e5660159039273df7e61c18050f267253086d89 +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-32/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..6ac66e15d5ebcd5ff208ec33180176c46e605c38 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f6b4aa56bd68d5b605dfbbb9695c820d5c012d1582b99a80bd56da5da652457 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..bcaad7482cc1813859d1e652f169fa68f16981e3 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:046455cb6ded9b2989f35bf5efbbff0517d6651771b52c93d31759906cafe479 +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..02f9ed8e0defc2ebe0d43c0d14abc27c32e2b2cf --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-32/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-32/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-32/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..f3bdacd3d072e0a631b89c4037b41375ef42fd59 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/tokens_state.json @@ -0,0 +1 @@ +{"total": 969264, "trainable": 14655} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a15832d78eb4d9b7f41431990a95ecf9956769a9 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + } + ], + "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": 6.578961181068442e+16, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-32/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-320/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-320/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-320/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b8e1e0278e8db8d629f21ace217798bfa36b8f0a --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:acefb37e7b13dc5d3275f02aadb2f638d6261fddc268c394ce7608c0d025d6df +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-320/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..92a887df53a65d0be5e93ca3066e15a0ce49dbd7 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5252e5c71f123b3511507288eed6bab62318023400dcbffeab20d27f0243691d +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..fb2d9796f065a223e6c4ea502c863112b2a15139 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0391605f7d7b8626c1c97d35086145a3e9e4df5cafc3842162a7664ce0f73f43 +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf4669ac28031073c15cff33e6a2bd7daa0e7337 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-320/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-320/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-320/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..421715dcdca9c91ad96c462d40a80afb871e72f5 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/tokens_state.json @@ -0,0 +1 @@ +{"total": 9678688, "trainable": 146668} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3fd3e379e715be6c20b93962a9619c0de4334749 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.03220707178115845, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.000319857761496678, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00032, + "step": 289, + "tokens/total": 8739680, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 132397 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.031606193631887436, + "learning_rate": 4.90663667927174e-05, + "loss": 0.0005622455500997603, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00056, + "step": 290, + "tokens/total": 8770112, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 132846 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.6305046677589417, + "learning_rate": 4.877873600900581e-05, + "loss": 0.006522824056446552, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00654, + "step": 291, + "tokens/total": 8800368, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 133347 + }, + { + "epoch": 1.140625, + "grad_norm": 0.03814006596803665, + "learning_rate": 4.849136411748306e-05, + "loss": 0.000703076075296849, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0007, + "step": 292, + "tokens/total": 8830688, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 133779 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.05072004348039627, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0011388614075258374, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00114, + "step": 293, + "tokens/total": 8861008, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 134251 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.2358824610710144, + "learning_rate": 4.791744483460251e-05, + "loss": 0.006398226134479046, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00642, + "step": 294, + "tokens/total": 8891760, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 134683 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2783265709877014, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0070259119383990765, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00705, + "step": 295, + "tokens/total": 8922192, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 135165 + }, + { + "epoch": 1.15625, + "grad_norm": 0.34820756316185, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.004574563354253769, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 296, + "tokens/total": 8952496, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 135599 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.03794313222169876, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0009930897504091263, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 297, + "tokens/total": 8982768, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 136038 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.19601035118103027, + "learning_rate": 4.677323837567412e-05, + "loss": 0.003028921550139785, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00303, + "step": 298, + "tokens/total": 9013248, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 136495 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0572715662419796, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.001314631663262844, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00132, + "step": 299, + "tokens/total": 9043632, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 136953 + }, + { + "epoch": 1.171875, + "grad_norm": 0.12685413658618927, + "learning_rate": 4.620314165804964e-05, + "loss": 0.0023695232812315226, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00237, + "step": 300, + "tokens/total": 9073936, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 137412 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.04133208841085434, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0008228939259424806, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00082, + "step": 301, + "tokens/total": 9104416, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 137907 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.01651018112897873, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0003794343792833388, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00038, + "step": 302, + "tokens/total": 9134688, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 138383 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.009257322177290916, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00024720095098018646, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00025, + "step": 303, + "tokens/total": 9165088, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 138842 + }, + { + "epoch": 1.1875, + "grad_norm": 0.038815200328826904, + "learning_rate": 4.506743570677743e-05, + "loss": 0.000565327238291502, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 304, + "tokens/total": 9195584, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 139269 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.37655678391456604, + "learning_rate": 4.478451305763618e-05, + "loss": 0.005295810755342245, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00531, + "step": 305, + "tokens/total": 9225760, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 139736 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.014935888350009918, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0003102960763499141, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00031, + "step": 306, + "tokens/total": 9256352, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 140226 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.02618814818561077, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0003336220979690552, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00033, + "step": 307, + "tokens/total": 9284256, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 140664 + }, + { + "epoch": 1.203125, + "grad_norm": 0.2529093027114868, + "learning_rate": 4.393834276419352e-05, + "loss": 0.005605834536254406, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00562, + "step": 308, + "tokens/total": 9314768, + "tokens/train_per_sec_per_gpu": 29.95, + "tokens/trainable": 141115 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0079645412042737, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00018522725440561771, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 309, + "tokens/total": 9344976, + "tokens/train_per_sec_per_gpu": 28.95, + "tokens/trainable": 141530 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.4591776132583618, + "learning_rate": 4.337651127546448e-05, + "loss": 0.003637854475528002, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00364, + "step": 310, + "tokens/total": 9375280, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 141964 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.02764446660876274, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00038255995605140924, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00038, + "step": 311, + "tokens/total": 9405536, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 142428 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0073210448026657104, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.0001625180011615157, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 312, + "tokens/total": 9435696, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 142894 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.1570003777742386, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00023247752687893808, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00023, + "step": 313, + "tokens/total": 9466080, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 143368 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.2202703058719635, + "learning_rate": 4.225874584473174e-05, + "loss": 0.002779184840619564, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00278, + "step": 314, + "tokens/total": 9496400, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 143824 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.04489271715283394, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0008375659235753119, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00084, + "step": 315, + "tokens/total": 9526880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 144306 + }, + { + "epoch": 1.234375, + "grad_norm": 0.006173362489789724, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001250750501640141, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 316, + "tokens/total": 9557280, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 144763 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.011352705769240856, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00022372312378138304, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 317, + "tokens/total": 9587712, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 145255 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.05640754476189613, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0007214375073090196, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00072, + "step": 318, + "tokens/total": 9618064, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 145714 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.15869443118572235, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0053039658814668655, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00532, + "step": 319, + "tokens/total": 9648336, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 146200 + }, + { + "epoch": 1.25, + "grad_norm": 0.44812121987342834, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0036084880121052265, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00362, + "step": 320, + "tokens/total": 9678688, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 146668 + } + ], + "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": 6.569491143349279e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-320/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-352/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-352/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-352/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a709e87c4a37839d95cfe1f46f93d653c959bcdf --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe94ef0ba516f83ec75c03ab2818b28cf78a7f79b5d42589c33734d4e96fce9f +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-352/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..a158c85bfd545a4d4d05c9e3207e39e25c828d11 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4268de87b83991938914f6bfbfe4e773e8cdfabde6548d3b0c2a73e730800c9f +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..0ae29c6e3264de0b6c52d65d7699ee19a45126df --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5d3595bca569f2399926c6c523fd98aead95fc6d00353c61841e5028ebbdeff +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..1d2b8d34c008174b230d6dbbee4469d934686b9f --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-352/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-352/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-352/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..639c485fed1f2598b14da4a28cd7a2ed978ce7cd --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/tokens_state.json @@ -0,0 +1 @@ +{"total": 10647760, "trainable": 161102} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c34502c2139901a6a0dd761af9c5e2f566a1d57f --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.03220707178115845, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.000319857761496678, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00032, + "step": 289, + "tokens/total": 8739680, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 132397 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.031606193631887436, + "learning_rate": 4.90663667927174e-05, + "loss": 0.0005622455500997603, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00056, + "step": 290, + "tokens/total": 8770112, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 132846 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.6305046677589417, + "learning_rate": 4.877873600900581e-05, + "loss": 0.006522824056446552, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00654, + "step": 291, + "tokens/total": 8800368, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 133347 + }, + { + "epoch": 1.140625, + "grad_norm": 0.03814006596803665, + "learning_rate": 4.849136411748306e-05, + "loss": 0.000703076075296849, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0007, + "step": 292, + "tokens/total": 8830688, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 133779 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.05072004348039627, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0011388614075258374, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00114, + "step": 293, + "tokens/total": 8861008, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 134251 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.2358824610710144, + "learning_rate": 4.791744483460251e-05, + "loss": 0.006398226134479046, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00642, + "step": 294, + "tokens/total": 8891760, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 134683 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2783265709877014, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0070259119383990765, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00705, + "step": 295, + "tokens/total": 8922192, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 135165 + }, + { + "epoch": 1.15625, + "grad_norm": 0.34820756316185, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.004574563354253769, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 296, + "tokens/total": 8952496, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 135599 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.03794313222169876, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0009930897504091263, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 297, + "tokens/total": 8982768, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 136038 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.19601035118103027, + "learning_rate": 4.677323837567412e-05, + "loss": 0.003028921550139785, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00303, + "step": 298, + "tokens/total": 9013248, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 136495 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0572715662419796, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.001314631663262844, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00132, + "step": 299, + "tokens/total": 9043632, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 136953 + }, + { + "epoch": 1.171875, + "grad_norm": 0.12685413658618927, + "learning_rate": 4.620314165804964e-05, + "loss": 0.0023695232812315226, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00237, + "step": 300, + "tokens/total": 9073936, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 137412 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.04133208841085434, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0008228939259424806, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00082, + "step": 301, + "tokens/total": 9104416, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 137907 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.01651018112897873, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0003794343792833388, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00038, + "step": 302, + "tokens/total": 9134688, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 138383 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.009257322177290916, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00024720095098018646, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00025, + "step": 303, + "tokens/total": 9165088, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 138842 + }, + { + "epoch": 1.1875, + "grad_norm": 0.038815200328826904, + "learning_rate": 4.506743570677743e-05, + "loss": 0.000565327238291502, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 304, + "tokens/total": 9195584, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 139269 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.37655678391456604, + "learning_rate": 4.478451305763618e-05, + "loss": 0.005295810755342245, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00531, + "step": 305, + "tokens/total": 9225760, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 139736 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.014935888350009918, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0003102960763499141, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00031, + "step": 306, + "tokens/total": 9256352, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 140226 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.02618814818561077, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0003336220979690552, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00033, + "step": 307, + "tokens/total": 9284256, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 140664 + }, + { + "epoch": 1.203125, + "grad_norm": 0.2529093027114868, + "learning_rate": 4.393834276419352e-05, + "loss": 0.005605834536254406, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00562, + "step": 308, + "tokens/total": 9314768, + "tokens/train_per_sec_per_gpu": 29.95, + "tokens/trainable": 141115 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0079645412042737, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00018522725440561771, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 309, + "tokens/total": 9344976, + "tokens/train_per_sec_per_gpu": 28.95, + "tokens/trainable": 141530 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.4591776132583618, + "learning_rate": 4.337651127546448e-05, + "loss": 0.003637854475528002, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00364, + "step": 310, + "tokens/total": 9375280, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 141964 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.02764446660876274, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00038255995605140924, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00038, + "step": 311, + "tokens/total": 9405536, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 142428 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0073210448026657104, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.0001625180011615157, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 312, + "tokens/total": 9435696, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 142894 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.1570003777742386, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00023247752687893808, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00023, + "step": 313, + "tokens/total": 9466080, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 143368 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.2202703058719635, + "learning_rate": 4.225874584473174e-05, + "loss": 0.002779184840619564, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00278, + "step": 314, + "tokens/total": 9496400, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 143824 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.04489271715283394, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0008375659235753119, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00084, + "step": 315, + "tokens/total": 9526880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 144306 + }, + { + "epoch": 1.234375, + "grad_norm": 0.006173362489789724, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001250750501640141, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 316, + "tokens/total": 9557280, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 144763 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.011352705769240856, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00022372312378138304, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 317, + "tokens/total": 9587712, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 145255 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.05640754476189613, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0007214375073090196, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00072, + "step": 318, + "tokens/total": 9618064, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 145714 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.15869443118572235, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0053039658814668655, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00532, + "step": 319, + "tokens/total": 9648336, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 146200 + }, + { + "epoch": 1.25, + "grad_norm": 0.44812121987342834, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0036084880121052265, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00362, + "step": 320, + "tokens/total": 9678688, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 146668 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.05665228143334389, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0009465725161135197, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00095, + "step": 321, + "tokens/total": 9708848, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 147114 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.016061583533883095, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00015383574645966291, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00015, + "step": 322, + "tokens/total": 9739360, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 147564 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.06347034126520157, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0009517542785033584, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00095, + "step": 323, + "tokens/total": 9769776, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 148013 + }, + { + "epoch": 1.265625, + "grad_norm": 0.24154019355773926, + "learning_rate": 3.95030593412612e-05, + "loss": 0.004087492823600769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.0041, + "step": 324, + "tokens/total": 9800096, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 148470 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.22498610615730286, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0027766539715230465, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00278, + "step": 325, + "tokens/total": 9830304, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 148926 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.002810331992805004, + "learning_rate": 3.895929566172861e-05, + "loss": 4.6118868340272456e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 326, + "tokens/total": 9860608, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 149420 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.007315841969102621, + "learning_rate": 3.868840945050728e-05, + "loss": 0.00011396597255952656, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00011, + "step": 327, + "tokens/total": 9890560, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 149873 + }, + { + "epoch": 1.28125, + "grad_norm": 0.021338002756237984, + "learning_rate": 3.841820203114995e-05, + "loss": 0.00021865982853341848, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 328, + "tokens/total": 9920880, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 150347 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.027939820662140846, + "learning_rate": 3.814868464809027e-05, + "loss": 0.0003473583492450416, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 329, + "tokens/total": 9951280, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 150822 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.0030000312253832817, + "learning_rate": 3.787986851704667e-05, + "loss": 5.043415876571089e-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": 9981600, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 151266 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.016975272446870804, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.0002860842505469918, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 331, + "tokens/total": 10012016, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 151734 + }, + { + "epoch": 1.296875, + "grad_norm": 0.0012887469492852688, + "learning_rate": 3.734438472750619e-05, + "loss": 2.801024675136432e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 332, + "tokens/total": 10042448, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 152184 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.0035721107851713896, + "learning_rate": 3.707773935267552e-05, + "loss": 4.7658577386755496e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 333, + "tokens/total": 10073024, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 152623 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.036152660846710205, + "learning_rate": 3.68118397962661e-05, + "loss": 0.00040513710700906813, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00041, + "step": 334, + "tokens/total": 10103504, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 153040 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.11260190606117249, + "learning_rate": 3.654669712344384e-05, + "loss": 0.0015010370407253504, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.0015, + "step": 335, + "tokens/total": 10131632, + "tokens/train_per_sec_per_gpu": 31.76, + "tokens/trainable": 153463 + }, + { + "epoch": 1.3125, + "grad_norm": 0.17602311074733734, + "learning_rate": 3.628232236787763e-05, + "loss": 0.007916431874036789, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00795, + "step": 336, + "tokens/total": 10161824, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 153908 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.005530365742743015, + "learning_rate": 3.6018726531280144e-05, + "loss": 8.388348942389712e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00008, + "step": 337, + "tokens/total": 10192448, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 154387 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.06029028072953224, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0002887388109229505, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00029, + "step": 338, + "tokens/total": 10222704, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 154836 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.003610784187912941, + "learning_rate": 3.549391545931585e-05, + "loss": 6.314883648883551e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 339, + "tokens/total": 10253168, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 155261 + }, + { + "epoch": 1.328125, + "grad_norm": 0.0029664942994713783, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.0910730604082346e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10283552, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 155674 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.3301459848880768, + "learning_rate": 3.49723512647657e-05, + "loss": 0.001420877524651587, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 341, + "tokens/total": 10313872, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 156164 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.0065552121959626675, + "learning_rate": 3.471281389826491e-05, + "loss": 3.5097506042802706e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00004, + "step": 342, + "tokens/total": 10344256, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 156611 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.027132108807563782, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.00014500679390039295, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00015, + "step": 343, + "tokens/total": 10374544, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 157073 + }, + { + "epoch": 1.34375, + "grad_norm": 0.40091755986213684, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.004070552997291088, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00408, + "step": 344, + "tokens/total": 10405040, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 157523 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.2770630419254303, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.005590756889432669, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00561, + "step": 345, + "tokens/total": 10435424, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 157964 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.007484348025172949, + "learning_rate": 3.3683214232914404e-05, + "loss": 8.245774370152503e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 346, + "tokens/total": 10465760, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 158390 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005284447688609362, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00010297571134287864, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 347, + "tokens/total": 10495904, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 158848 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0784892812371254, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0006088463123887777, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00061, + "step": 348, + "tokens/total": 10526624, + "tokens/train_per_sec_per_gpu": 28.7, + "tokens/trainable": 159270 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.33535251021385193, + "learning_rate": 3.292029118616024e-05, + "loss": 0.006133331451565027, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00615, + "step": 349, + "tokens/total": 10556752, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 159720 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.008170238696038723, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00012147967208875343, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00012, + "step": 350, + "tokens/total": 10587264, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 160187 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.001569622429087758, + "learning_rate": 3.241625231705354e-05, + "loss": 3.8519159716088325e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 351, + "tokens/total": 10617504, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 160645 + }, + { + "epoch": 1.375, + "grad_norm": 0.143654003739357, + "learning_rate": 3.216563735127618e-05, + "loss": 0.002758543938398361, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00276, + "step": 352, + "tokens/total": 10647760, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 161102 + } + ], + "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": 7.227256939836134e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-352/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-384/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-384/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-384/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..d336be2835dcba0f3a78875a4c65063e545b6e90 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9de877f9666be8b53ab2945a7578d5435443eee3aa2df38a704a119bd505fc94 +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-384/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..51710496cca874ea8f3acea9919e24f04d37c933 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20142ea7876ad64dae3845ba642abaee2ddb63d1860061e881e5fb77b004aaa1 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..5aea99e5163fa108032b2b18ebc587f398e76e51 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea55d7f6de0215cad74ef9143c41f13fa0fffc3e1c9c7025b8d1c96ebbb6f69e +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c676299e10f0344839a2b08f1cc1e004410cac0 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-384/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-384/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-384/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..de0b107094014b7da8157ab906e76603a905fde8 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/tokens_state.json @@ -0,0 +1 @@ +{"total": 11618640, "trainable": 175720} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c2fd8398d025ef645b6df49050b9ee71c2fbbfda --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.03220707178115845, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.000319857761496678, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00032, + "step": 289, + "tokens/total": 8739680, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 132397 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.031606193631887436, + "learning_rate": 4.90663667927174e-05, + "loss": 0.0005622455500997603, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00056, + "step": 290, + "tokens/total": 8770112, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 132846 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.6305046677589417, + "learning_rate": 4.877873600900581e-05, + "loss": 0.006522824056446552, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00654, + "step": 291, + "tokens/total": 8800368, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 133347 + }, + { + "epoch": 1.140625, + "grad_norm": 0.03814006596803665, + "learning_rate": 4.849136411748306e-05, + "loss": 0.000703076075296849, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0007, + "step": 292, + "tokens/total": 8830688, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 133779 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.05072004348039627, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0011388614075258374, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00114, + "step": 293, + "tokens/total": 8861008, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 134251 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.2358824610710144, + "learning_rate": 4.791744483460251e-05, + "loss": 0.006398226134479046, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00642, + "step": 294, + "tokens/total": 8891760, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 134683 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2783265709877014, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0070259119383990765, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00705, + "step": 295, + "tokens/total": 8922192, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 135165 + }, + { + "epoch": 1.15625, + "grad_norm": 0.34820756316185, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.004574563354253769, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 296, + "tokens/total": 8952496, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 135599 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.03794313222169876, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0009930897504091263, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 297, + "tokens/total": 8982768, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 136038 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.19601035118103027, + "learning_rate": 4.677323837567412e-05, + "loss": 0.003028921550139785, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00303, + "step": 298, + "tokens/total": 9013248, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 136495 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0572715662419796, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.001314631663262844, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00132, + "step": 299, + "tokens/total": 9043632, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 136953 + }, + { + "epoch": 1.171875, + "grad_norm": 0.12685413658618927, + "learning_rate": 4.620314165804964e-05, + "loss": 0.0023695232812315226, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00237, + "step": 300, + "tokens/total": 9073936, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 137412 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.04133208841085434, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0008228939259424806, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00082, + "step": 301, + "tokens/total": 9104416, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 137907 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.01651018112897873, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0003794343792833388, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00038, + "step": 302, + "tokens/total": 9134688, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 138383 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.009257322177290916, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00024720095098018646, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00025, + "step": 303, + "tokens/total": 9165088, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 138842 + }, + { + "epoch": 1.1875, + "grad_norm": 0.038815200328826904, + "learning_rate": 4.506743570677743e-05, + "loss": 0.000565327238291502, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 304, + "tokens/total": 9195584, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 139269 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.37655678391456604, + "learning_rate": 4.478451305763618e-05, + "loss": 0.005295810755342245, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00531, + "step": 305, + "tokens/total": 9225760, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 139736 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.014935888350009918, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0003102960763499141, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00031, + "step": 306, + "tokens/total": 9256352, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 140226 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.02618814818561077, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0003336220979690552, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00033, + "step": 307, + "tokens/total": 9284256, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 140664 + }, + { + "epoch": 1.203125, + "grad_norm": 0.2529093027114868, + "learning_rate": 4.393834276419352e-05, + "loss": 0.005605834536254406, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00562, + "step": 308, + "tokens/total": 9314768, + "tokens/train_per_sec_per_gpu": 29.95, + "tokens/trainable": 141115 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0079645412042737, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00018522725440561771, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 309, + "tokens/total": 9344976, + "tokens/train_per_sec_per_gpu": 28.95, + "tokens/trainable": 141530 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.4591776132583618, + "learning_rate": 4.337651127546448e-05, + "loss": 0.003637854475528002, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00364, + "step": 310, + "tokens/total": 9375280, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 141964 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.02764446660876274, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00038255995605140924, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00038, + "step": 311, + "tokens/total": 9405536, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 142428 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0073210448026657104, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.0001625180011615157, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 312, + "tokens/total": 9435696, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 142894 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.1570003777742386, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00023247752687893808, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00023, + "step": 313, + "tokens/total": 9466080, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 143368 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.2202703058719635, + "learning_rate": 4.225874584473174e-05, + "loss": 0.002779184840619564, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00278, + "step": 314, + "tokens/total": 9496400, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 143824 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.04489271715283394, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0008375659235753119, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00084, + "step": 315, + "tokens/total": 9526880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 144306 + }, + { + "epoch": 1.234375, + "grad_norm": 0.006173362489789724, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001250750501640141, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 316, + "tokens/total": 9557280, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 144763 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.011352705769240856, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00022372312378138304, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 317, + "tokens/total": 9587712, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 145255 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.05640754476189613, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0007214375073090196, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00072, + "step": 318, + "tokens/total": 9618064, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 145714 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.15869443118572235, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0053039658814668655, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00532, + "step": 319, + "tokens/total": 9648336, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 146200 + }, + { + "epoch": 1.25, + "grad_norm": 0.44812121987342834, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0036084880121052265, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00362, + "step": 320, + "tokens/total": 9678688, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 146668 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.05665228143334389, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0009465725161135197, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00095, + "step": 321, + "tokens/total": 9708848, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 147114 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.016061583533883095, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00015383574645966291, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00015, + "step": 322, + "tokens/total": 9739360, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 147564 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.06347034126520157, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0009517542785033584, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00095, + "step": 323, + "tokens/total": 9769776, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 148013 + }, + { + "epoch": 1.265625, + "grad_norm": 0.24154019355773926, + "learning_rate": 3.95030593412612e-05, + "loss": 0.004087492823600769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.0041, + "step": 324, + "tokens/total": 9800096, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 148470 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.22498610615730286, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0027766539715230465, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00278, + "step": 325, + "tokens/total": 9830304, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 148926 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.002810331992805004, + "learning_rate": 3.895929566172861e-05, + "loss": 4.6118868340272456e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 326, + "tokens/total": 9860608, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 149420 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.007315841969102621, + "learning_rate": 3.868840945050728e-05, + "loss": 0.00011396597255952656, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00011, + "step": 327, + "tokens/total": 9890560, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 149873 + }, + { + "epoch": 1.28125, + "grad_norm": 0.021338002756237984, + "learning_rate": 3.841820203114995e-05, + "loss": 0.00021865982853341848, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 328, + "tokens/total": 9920880, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 150347 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.027939820662140846, + "learning_rate": 3.814868464809027e-05, + "loss": 0.0003473583492450416, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 329, + "tokens/total": 9951280, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 150822 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.0030000312253832817, + "learning_rate": 3.787986851704667e-05, + "loss": 5.043415876571089e-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": 9981600, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 151266 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.016975272446870804, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.0002860842505469918, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 331, + "tokens/total": 10012016, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 151734 + }, + { + "epoch": 1.296875, + "grad_norm": 0.0012887469492852688, + "learning_rate": 3.734438472750619e-05, + "loss": 2.801024675136432e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 332, + "tokens/total": 10042448, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 152184 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.0035721107851713896, + "learning_rate": 3.707773935267552e-05, + "loss": 4.7658577386755496e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 333, + "tokens/total": 10073024, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 152623 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.036152660846710205, + "learning_rate": 3.68118397962661e-05, + "loss": 0.00040513710700906813, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00041, + "step": 334, + "tokens/total": 10103504, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 153040 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.11260190606117249, + "learning_rate": 3.654669712344384e-05, + "loss": 0.0015010370407253504, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.0015, + "step": 335, + "tokens/total": 10131632, + "tokens/train_per_sec_per_gpu": 31.76, + "tokens/trainable": 153463 + }, + { + "epoch": 1.3125, + "grad_norm": 0.17602311074733734, + "learning_rate": 3.628232236787763e-05, + "loss": 0.007916431874036789, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00795, + "step": 336, + "tokens/total": 10161824, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 153908 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.005530365742743015, + "learning_rate": 3.6018726531280144e-05, + "loss": 8.388348942389712e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00008, + "step": 337, + "tokens/total": 10192448, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 154387 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.06029028072953224, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0002887388109229505, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00029, + "step": 338, + "tokens/total": 10222704, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 154836 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.003610784187912941, + "learning_rate": 3.549391545931585e-05, + "loss": 6.314883648883551e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 339, + "tokens/total": 10253168, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 155261 + }, + { + "epoch": 1.328125, + "grad_norm": 0.0029664942994713783, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.0910730604082346e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10283552, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 155674 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.3301459848880768, + "learning_rate": 3.49723512647657e-05, + "loss": 0.001420877524651587, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 341, + "tokens/total": 10313872, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 156164 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.0065552121959626675, + "learning_rate": 3.471281389826491e-05, + "loss": 3.5097506042802706e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00004, + "step": 342, + "tokens/total": 10344256, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 156611 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.027132108807563782, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.00014500679390039295, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00015, + "step": 343, + "tokens/total": 10374544, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 157073 + }, + { + "epoch": 1.34375, + "grad_norm": 0.40091755986213684, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.004070552997291088, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00408, + "step": 344, + "tokens/total": 10405040, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 157523 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.2770630419254303, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.005590756889432669, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00561, + "step": 345, + "tokens/total": 10435424, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 157964 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.007484348025172949, + "learning_rate": 3.3683214232914404e-05, + "loss": 8.245774370152503e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 346, + "tokens/total": 10465760, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 158390 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005284447688609362, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00010297571134287864, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 347, + "tokens/total": 10495904, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 158848 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0784892812371254, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0006088463123887777, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00061, + "step": 348, + "tokens/total": 10526624, + "tokens/train_per_sec_per_gpu": 28.7, + "tokens/trainable": 159270 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.33535251021385193, + "learning_rate": 3.292029118616024e-05, + "loss": 0.006133331451565027, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00615, + "step": 349, + "tokens/total": 10556752, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 159720 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.008170238696038723, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00012147967208875343, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00012, + "step": 350, + "tokens/total": 10587264, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 160187 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.001569622429087758, + "learning_rate": 3.241625231705354e-05, + "loss": 3.8519159716088325e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 351, + "tokens/total": 10617504, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 160645 + }, + { + "epoch": 1.375, + "grad_norm": 0.143654003739357, + "learning_rate": 3.216563735127618e-05, + "loss": 0.002758543938398361, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00276, + "step": 352, + "tokens/total": 10647760, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 161102 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.042433012276887894, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00035650472273118794, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 353, + "tokens/total": 10678080, + "tokens/train_per_sec_per_gpu": 28.87, + "tokens/trainable": 161555 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.9453999996185303, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0030506744515150785, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00306, + "step": 354, + "tokens/total": 10708544, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 162020 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.03272778168320656, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00023931957548484206, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00024, + "step": 355, + "tokens/total": 10739024, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 162448 + }, + { + "epoch": 1.390625, + "grad_norm": 0.006854584906250238, + "learning_rate": 3.11727834939056e-05, + "loss": 6.610051059396937e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10769024, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 162863 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.03258929029107094, + "learning_rate": 3.092702317708967e-05, + "loss": 0.0002595578844193369, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00026, + "step": 357, + "tokens/total": 10799328, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 163318 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.21498654782772064, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.003778135636821389, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00379, + "step": 358, + "tokens/total": 10829488, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 163767 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.22063465416431427, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.00428510969504714, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00429, + "step": 359, + "tokens/total": 10859936, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 164208 + }, + { + "epoch": 1.40625, + "grad_norm": 0.01306198537349701, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00016752032388467342, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00017, + "step": 360, + "tokens/total": 10890400, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 164655 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.42359107732772827, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.022302038967609406, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02255, + "step": 361, + "tokens/total": 10920624, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 165100 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.14819112420082092, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.002135487971827388, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00214, + "step": 362, + "tokens/total": 10951136, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 165587 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.10172182321548462, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.0012491067172959447, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00125, + "step": 363, + "tokens/total": 10981344, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 166034 + }, + { + "epoch": 1.421875, + "grad_norm": 0.16598454117774963, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0014915807405486703, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00149, + "step": 364, + "tokens/total": 11011552, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 166485 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.01892480067908764, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00023381829669233412, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00023, + "step": 365, + "tokens/total": 11041872, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 166973 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.050091467797756195, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00035884848330169916, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00036, + "step": 366, + "tokens/total": 11072192, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 167414 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.6703732013702393, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.008329304866492748, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00836, + "step": 367, + "tokens/total": 11102736, + "tokens/train_per_sec_per_gpu": 30.18, + "tokens/trainable": 167851 + }, + { + "epoch": 1.4375, + "grad_norm": 0.10087089240550995, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0014869305305182934, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00149, + "step": 368, + "tokens/total": 11133056, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 168294 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.03148820623755455, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00043267637374810874, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 369, + "tokens/total": 11163568, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 168769 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.01531990710645914, + "learning_rate": 2.782696506053033e-05, + "loss": 0.00021343299886211753, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00021, + "step": 370, + "tokens/total": 11193936, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 169270 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.006444540806114674, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.00014627019118051976, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00015, + "step": 371, + "tokens/total": 11224480, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 169711 + }, + { + "epoch": 1.453125, + "grad_norm": 0.31427323818206787, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.00462032575160265, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00463, + "step": 372, + "tokens/total": 11254912, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 170185 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.013480601832270622, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.000224081173655577, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00022, + "step": 373, + "tokens/total": 11285104, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 170650 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.023016467690467834, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0001827064115786925, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00018, + "step": 374, + "tokens/total": 11315600, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 171126 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.019638920202851295, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0005156319821253419, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00052, + "step": 375, + "tokens/total": 11345776, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 171600 + }, + { + "epoch": 1.46875, + "grad_norm": 0.03251553326845169, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004913917509838939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00049, + "step": 376, + "tokens/total": 11376208, + "tokens/train_per_sec_per_gpu": 28.1, + "tokens/trainable": 172017 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.03556421399116516, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0005220613675191998, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00052, + "step": 377, + "tokens/total": 11406512, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 172493 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.04202103987336159, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0004697944095823914, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00047, + "step": 378, + "tokens/total": 11436544, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 172951 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.023767679929733276, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0004130054439883679, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00041, + "step": 379, + "tokens/total": 11467008, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 173435 + }, + { + "epoch": 1.484375, + "grad_norm": 0.04106076434254646, + "learning_rate": 2.557117581955798e-05, + "loss": 0.00047382546472363174, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00047, + "step": 380, + "tokens/total": 11497184, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 173900 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.15864700078964233, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.003480169689282775, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00349, + "step": 381, + "tokens/total": 11527600, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 174371 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.04146522656083107, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0005609758081845939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00056, + "step": 382, + "tokens/total": 11557808, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 174826 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.012171853333711624, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00022788115893490613, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00023, + "step": 383, + "tokens/total": 11588352, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 175276 + }, + { + "epoch": 1.5, + "grad_norm": 0.08181510865688324, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0016442432533949614, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 384, + "tokens/total": 11618640, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 175720 + } + ], + "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": 7.886249931577882e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-384/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-416/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-416/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-416/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c3967390e1258a01e83a396e8061b2130cdf5f2f --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34249439ede5decaab8e3dd5072ee329f2fbdc84afa50d247ddbf85397ad24f6 +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-416/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..e428c940c9130af188b1b0c09e9b74ad8d547c51 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cdb90f7e19d962cf814644ff9467861be2a01b30d9c66d67fa82787892691aa3 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..323da0a5194e5ce925b57124831414c61390f5e0 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49f752bd9a9a6e4604278f4cbc2ad9ae381068763e6148938859abe0b809968f +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..20a981d8c8f57b8d2e18429ceb299414229ad6ac --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-416/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-416/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-416/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..141e4b902aa713abf9a208543b41cda2cb5debfc --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/tokens_state.json @@ -0,0 +1 @@ +{"total": 12586736, "trainable": 190216} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..51c06ffc08eee4351f8eb0d20d6b7d42dd7d53d8 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.03220707178115845, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.000319857761496678, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00032, + "step": 289, + "tokens/total": 8739680, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 132397 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.031606193631887436, + "learning_rate": 4.90663667927174e-05, + "loss": 0.0005622455500997603, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00056, + "step": 290, + "tokens/total": 8770112, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 132846 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.6305046677589417, + "learning_rate": 4.877873600900581e-05, + "loss": 0.006522824056446552, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00654, + "step": 291, + "tokens/total": 8800368, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 133347 + }, + { + "epoch": 1.140625, + "grad_norm": 0.03814006596803665, + "learning_rate": 4.849136411748306e-05, + "loss": 0.000703076075296849, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0007, + "step": 292, + "tokens/total": 8830688, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 133779 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.05072004348039627, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0011388614075258374, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00114, + "step": 293, + "tokens/total": 8861008, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 134251 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.2358824610710144, + "learning_rate": 4.791744483460251e-05, + "loss": 0.006398226134479046, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00642, + "step": 294, + "tokens/total": 8891760, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 134683 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2783265709877014, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0070259119383990765, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00705, + "step": 295, + "tokens/total": 8922192, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 135165 + }, + { + "epoch": 1.15625, + "grad_norm": 0.34820756316185, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.004574563354253769, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 296, + "tokens/total": 8952496, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 135599 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.03794313222169876, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0009930897504091263, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 297, + "tokens/total": 8982768, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 136038 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.19601035118103027, + "learning_rate": 4.677323837567412e-05, + "loss": 0.003028921550139785, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00303, + "step": 298, + "tokens/total": 9013248, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 136495 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0572715662419796, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.001314631663262844, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00132, + "step": 299, + "tokens/total": 9043632, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 136953 + }, + { + "epoch": 1.171875, + "grad_norm": 0.12685413658618927, + "learning_rate": 4.620314165804964e-05, + "loss": 0.0023695232812315226, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00237, + "step": 300, + "tokens/total": 9073936, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 137412 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.04133208841085434, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0008228939259424806, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00082, + "step": 301, + "tokens/total": 9104416, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 137907 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.01651018112897873, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0003794343792833388, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00038, + "step": 302, + "tokens/total": 9134688, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 138383 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.009257322177290916, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00024720095098018646, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00025, + "step": 303, + "tokens/total": 9165088, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 138842 + }, + { + "epoch": 1.1875, + "grad_norm": 0.038815200328826904, + "learning_rate": 4.506743570677743e-05, + "loss": 0.000565327238291502, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 304, + "tokens/total": 9195584, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 139269 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.37655678391456604, + "learning_rate": 4.478451305763618e-05, + "loss": 0.005295810755342245, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00531, + "step": 305, + "tokens/total": 9225760, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 139736 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.014935888350009918, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0003102960763499141, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00031, + "step": 306, + "tokens/total": 9256352, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 140226 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.02618814818561077, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0003336220979690552, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00033, + "step": 307, + "tokens/total": 9284256, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 140664 + }, + { + "epoch": 1.203125, + "grad_norm": 0.2529093027114868, + "learning_rate": 4.393834276419352e-05, + "loss": 0.005605834536254406, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00562, + "step": 308, + "tokens/total": 9314768, + "tokens/train_per_sec_per_gpu": 29.95, + "tokens/trainable": 141115 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0079645412042737, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00018522725440561771, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 309, + "tokens/total": 9344976, + "tokens/train_per_sec_per_gpu": 28.95, + "tokens/trainable": 141530 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.4591776132583618, + "learning_rate": 4.337651127546448e-05, + "loss": 0.003637854475528002, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00364, + "step": 310, + "tokens/total": 9375280, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 141964 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.02764446660876274, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00038255995605140924, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00038, + "step": 311, + "tokens/total": 9405536, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 142428 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0073210448026657104, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.0001625180011615157, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 312, + "tokens/total": 9435696, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 142894 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.1570003777742386, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00023247752687893808, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00023, + "step": 313, + "tokens/total": 9466080, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 143368 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.2202703058719635, + "learning_rate": 4.225874584473174e-05, + "loss": 0.002779184840619564, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00278, + "step": 314, + "tokens/total": 9496400, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 143824 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.04489271715283394, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0008375659235753119, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00084, + "step": 315, + "tokens/total": 9526880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 144306 + }, + { + "epoch": 1.234375, + "grad_norm": 0.006173362489789724, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001250750501640141, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 316, + "tokens/total": 9557280, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 144763 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.011352705769240856, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00022372312378138304, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 317, + "tokens/total": 9587712, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 145255 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.05640754476189613, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0007214375073090196, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00072, + "step": 318, + "tokens/total": 9618064, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 145714 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.15869443118572235, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0053039658814668655, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00532, + "step": 319, + "tokens/total": 9648336, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 146200 + }, + { + "epoch": 1.25, + "grad_norm": 0.44812121987342834, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0036084880121052265, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00362, + "step": 320, + "tokens/total": 9678688, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 146668 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.05665228143334389, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0009465725161135197, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00095, + "step": 321, + "tokens/total": 9708848, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 147114 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.016061583533883095, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00015383574645966291, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00015, + "step": 322, + "tokens/total": 9739360, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 147564 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.06347034126520157, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0009517542785033584, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00095, + "step": 323, + "tokens/total": 9769776, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 148013 + }, + { + "epoch": 1.265625, + "grad_norm": 0.24154019355773926, + "learning_rate": 3.95030593412612e-05, + "loss": 0.004087492823600769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.0041, + "step": 324, + "tokens/total": 9800096, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 148470 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.22498610615730286, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0027766539715230465, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00278, + "step": 325, + "tokens/total": 9830304, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 148926 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.002810331992805004, + "learning_rate": 3.895929566172861e-05, + "loss": 4.6118868340272456e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 326, + "tokens/total": 9860608, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 149420 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.007315841969102621, + "learning_rate": 3.868840945050728e-05, + "loss": 0.00011396597255952656, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00011, + "step": 327, + "tokens/total": 9890560, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 149873 + }, + { + "epoch": 1.28125, + "grad_norm": 0.021338002756237984, + "learning_rate": 3.841820203114995e-05, + "loss": 0.00021865982853341848, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 328, + "tokens/total": 9920880, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 150347 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.027939820662140846, + "learning_rate": 3.814868464809027e-05, + "loss": 0.0003473583492450416, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 329, + "tokens/total": 9951280, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 150822 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.0030000312253832817, + "learning_rate": 3.787986851704667e-05, + "loss": 5.043415876571089e-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": 9981600, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 151266 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.016975272446870804, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.0002860842505469918, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 331, + "tokens/total": 10012016, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 151734 + }, + { + "epoch": 1.296875, + "grad_norm": 0.0012887469492852688, + "learning_rate": 3.734438472750619e-05, + "loss": 2.801024675136432e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 332, + "tokens/total": 10042448, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 152184 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.0035721107851713896, + "learning_rate": 3.707773935267552e-05, + "loss": 4.7658577386755496e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 333, + "tokens/total": 10073024, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 152623 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.036152660846710205, + "learning_rate": 3.68118397962661e-05, + "loss": 0.00040513710700906813, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00041, + "step": 334, + "tokens/total": 10103504, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 153040 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.11260190606117249, + "learning_rate": 3.654669712344384e-05, + "loss": 0.0015010370407253504, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.0015, + "step": 335, + "tokens/total": 10131632, + "tokens/train_per_sec_per_gpu": 31.76, + "tokens/trainable": 153463 + }, + { + "epoch": 1.3125, + "grad_norm": 0.17602311074733734, + "learning_rate": 3.628232236787763e-05, + "loss": 0.007916431874036789, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00795, + "step": 336, + "tokens/total": 10161824, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 153908 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.005530365742743015, + "learning_rate": 3.6018726531280144e-05, + "loss": 8.388348942389712e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00008, + "step": 337, + "tokens/total": 10192448, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 154387 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.06029028072953224, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0002887388109229505, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00029, + "step": 338, + "tokens/total": 10222704, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 154836 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.003610784187912941, + "learning_rate": 3.549391545931585e-05, + "loss": 6.314883648883551e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 339, + "tokens/total": 10253168, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 155261 + }, + { + "epoch": 1.328125, + "grad_norm": 0.0029664942994713783, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.0910730604082346e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10283552, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 155674 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.3301459848880768, + "learning_rate": 3.49723512647657e-05, + "loss": 0.001420877524651587, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 341, + "tokens/total": 10313872, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 156164 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.0065552121959626675, + "learning_rate": 3.471281389826491e-05, + "loss": 3.5097506042802706e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00004, + "step": 342, + "tokens/total": 10344256, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 156611 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.027132108807563782, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.00014500679390039295, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00015, + "step": 343, + "tokens/total": 10374544, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 157073 + }, + { + "epoch": 1.34375, + "grad_norm": 0.40091755986213684, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.004070552997291088, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00408, + "step": 344, + "tokens/total": 10405040, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 157523 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.2770630419254303, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.005590756889432669, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00561, + "step": 345, + "tokens/total": 10435424, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 157964 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.007484348025172949, + "learning_rate": 3.3683214232914404e-05, + "loss": 8.245774370152503e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 346, + "tokens/total": 10465760, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 158390 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005284447688609362, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00010297571134287864, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 347, + "tokens/total": 10495904, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 158848 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0784892812371254, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0006088463123887777, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00061, + "step": 348, + "tokens/total": 10526624, + "tokens/train_per_sec_per_gpu": 28.7, + "tokens/trainable": 159270 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.33535251021385193, + "learning_rate": 3.292029118616024e-05, + "loss": 0.006133331451565027, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00615, + "step": 349, + "tokens/total": 10556752, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 159720 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.008170238696038723, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00012147967208875343, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00012, + "step": 350, + "tokens/total": 10587264, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 160187 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.001569622429087758, + "learning_rate": 3.241625231705354e-05, + "loss": 3.8519159716088325e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 351, + "tokens/total": 10617504, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 160645 + }, + { + "epoch": 1.375, + "grad_norm": 0.143654003739357, + "learning_rate": 3.216563735127618e-05, + "loss": 0.002758543938398361, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00276, + "step": 352, + "tokens/total": 10647760, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 161102 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.042433012276887894, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00035650472273118794, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 353, + "tokens/total": 10678080, + "tokens/train_per_sec_per_gpu": 28.87, + "tokens/trainable": 161555 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.9453999996185303, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0030506744515150785, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00306, + "step": 354, + "tokens/total": 10708544, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 162020 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.03272778168320656, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00023931957548484206, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00024, + "step": 355, + "tokens/total": 10739024, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 162448 + }, + { + "epoch": 1.390625, + "grad_norm": 0.006854584906250238, + "learning_rate": 3.11727834939056e-05, + "loss": 6.610051059396937e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10769024, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 162863 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.03258929029107094, + "learning_rate": 3.092702317708967e-05, + "loss": 0.0002595578844193369, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00026, + "step": 357, + "tokens/total": 10799328, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 163318 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.21498654782772064, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.003778135636821389, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00379, + "step": 358, + "tokens/total": 10829488, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 163767 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.22063465416431427, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.00428510969504714, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00429, + "step": 359, + "tokens/total": 10859936, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 164208 + }, + { + "epoch": 1.40625, + "grad_norm": 0.01306198537349701, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00016752032388467342, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00017, + "step": 360, + "tokens/total": 10890400, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 164655 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.42359107732772827, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.022302038967609406, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02255, + "step": 361, + "tokens/total": 10920624, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 165100 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.14819112420082092, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.002135487971827388, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00214, + "step": 362, + "tokens/total": 10951136, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 165587 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.10172182321548462, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.0012491067172959447, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00125, + "step": 363, + "tokens/total": 10981344, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 166034 + }, + { + "epoch": 1.421875, + "grad_norm": 0.16598454117774963, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0014915807405486703, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00149, + "step": 364, + "tokens/total": 11011552, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 166485 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.01892480067908764, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00023381829669233412, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00023, + "step": 365, + "tokens/total": 11041872, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 166973 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.050091467797756195, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00035884848330169916, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00036, + "step": 366, + "tokens/total": 11072192, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 167414 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.6703732013702393, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.008329304866492748, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00836, + "step": 367, + "tokens/total": 11102736, + "tokens/train_per_sec_per_gpu": 30.18, + "tokens/trainable": 167851 + }, + { + "epoch": 1.4375, + "grad_norm": 0.10087089240550995, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0014869305305182934, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00149, + "step": 368, + "tokens/total": 11133056, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 168294 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.03148820623755455, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00043267637374810874, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 369, + "tokens/total": 11163568, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 168769 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.01531990710645914, + "learning_rate": 2.782696506053033e-05, + "loss": 0.00021343299886211753, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00021, + "step": 370, + "tokens/total": 11193936, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 169270 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.006444540806114674, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.00014627019118051976, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00015, + "step": 371, + "tokens/total": 11224480, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 169711 + }, + { + "epoch": 1.453125, + "grad_norm": 0.31427323818206787, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.00462032575160265, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00463, + "step": 372, + "tokens/total": 11254912, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 170185 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.013480601832270622, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.000224081173655577, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00022, + "step": 373, + "tokens/total": 11285104, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 170650 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.023016467690467834, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0001827064115786925, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00018, + "step": 374, + "tokens/total": 11315600, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 171126 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.019638920202851295, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0005156319821253419, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00052, + "step": 375, + "tokens/total": 11345776, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 171600 + }, + { + "epoch": 1.46875, + "grad_norm": 0.03251553326845169, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004913917509838939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00049, + "step": 376, + "tokens/total": 11376208, + "tokens/train_per_sec_per_gpu": 28.1, + "tokens/trainable": 172017 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.03556421399116516, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0005220613675191998, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00052, + "step": 377, + "tokens/total": 11406512, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 172493 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.04202103987336159, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0004697944095823914, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00047, + "step": 378, + "tokens/total": 11436544, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 172951 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.023767679929733276, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0004130054439883679, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00041, + "step": 379, + "tokens/total": 11467008, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 173435 + }, + { + "epoch": 1.484375, + "grad_norm": 0.04106076434254646, + "learning_rate": 2.557117581955798e-05, + "loss": 0.00047382546472363174, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00047, + "step": 380, + "tokens/total": 11497184, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 173900 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.15864700078964233, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.003480169689282775, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00349, + "step": 381, + "tokens/total": 11527600, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 174371 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.04146522656083107, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0005609758081845939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00056, + "step": 382, + "tokens/total": 11557808, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 174826 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.012171853333711624, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00022788115893490613, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00023, + "step": 383, + "tokens/total": 11588352, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 175276 + }, + { + "epoch": 1.5, + "grad_norm": 0.08181510865688324, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0016442432533949614, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 384, + "tokens/total": 11618640, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 175720 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.10246019810438156, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0017187006305903196, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00172, + "step": 385, + "tokens/total": 11649072, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 176203 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.2453320175409317, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0038096809294074774, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00382, + "step": 386, + "tokens/total": 11679552, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 176651 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.005247004795819521, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00010425677464809269, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 387, + "tokens/total": 11709760, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 177115 + }, + { + "epoch": 1.515625, + "grad_norm": 0.027557479217648506, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00036113697569817305, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00036, + "step": 388, + "tokens/total": 11740288, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 177601 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.028131073340773582, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0005047487211413682, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0005, + "step": 389, + "tokens/total": 11770592, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 178050 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.0033069071359932423, + "learning_rate": 2.3437809889624914e-05, + "loss": 7.658063259441406e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00008, + "step": 390, + "tokens/total": 11800752, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 178531 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.028486015275120735, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0005519004189409316, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00055, + "step": 391, + "tokens/total": 11831360, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 179000 + }, + { + "epoch": 1.53125, + "grad_norm": 0.002584393136203289, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.4768250265624374e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 392, + "tokens/total": 11861552, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 179457 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.009344775229692459, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.0002210603270214051, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00022, + "step": 393, + "tokens/total": 11891904, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 179890 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.004884887952357531, + "learning_rate": 2.2620747959533722e-05, + "loss": 8.428651199210435e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00008, + "step": 394, + "tokens/total": 11922208, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 180341 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.023302890360355377, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.000332253344822675, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00033, + "step": 395, + "tokens/total": 11952672, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 180762 + }, + { + "epoch": 1.546875, + "grad_norm": 0.14256854355335236, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00286297220736742, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00287, + "step": 396, + "tokens/total": 11983088, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 181226 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.00555575592443347, + "learning_rate": 2.202206960760984e-05, + "loss": 9.771026816451922e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0001, + "step": 397, + "tokens/total": 12013488, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 181714 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.007696876302361488, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0001580697571625933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00016, + "step": 398, + "tokens/total": 12043968, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 182142 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.0037960875779390335, + "learning_rate": 2.1629798596457056e-05, + "loss": 7.467059913324192e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 399, + "tokens/total": 12074160, + "tokens/train_per_sec_per_gpu": 34.82, + "tokens/trainable": 182604 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004514187108725309, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.00011837383499369025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00012, + "step": 400, + "tokens/total": 12104352, + "tokens/train_per_sec_per_gpu": 25.26, + "tokens/trainable": 183011 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.09468130767345428, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0007745609618723392, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00077, + "step": 401, + "tokens/total": 12134240, + "tokens/train_per_sec_per_gpu": 31.75, + "tokens/trainable": 183459 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.017941495403647423, + "learning_rate": 2.105182715082638e-05, + "loss": 0.00024286848201882094, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 402, + "tokens/total": 12164480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 183959 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.005621996242552996, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.00012180282647022977, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00012, + "step": 403, + "tokens/total": 12194640, + "tokens/train_per_sec_per_gpu": 32.46, + "tokens/trainable": 184431 + }, + { + "epoch": 1.578125, + "grad_norm": 0.004712392576038837, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010195293725701049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 404, + "tokens/total": 12224960, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 184867 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.023734599351882935, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.00040538967004977167, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00041, + "step": 405, + "tokens/total": 12255008, + "tokens/train_per_sec_per_gpu": 29.33, + "tokens/trainable": 185300 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.06495791673660278, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.000537894491571933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00054, + "step": 406, + "tokens/total": 12285248, + "tokens/train_per_sec_per_gpu": 31.47, + "tokens/trainable": 185754 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.003490478964522481, + "learning_rate": 2.011689980574966e-05, + "loss": 6.170629058033228e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00006, + "step": 407, + "tokens/total": 12315552, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 186213 + }, + { + "epoch": 1.59375, + "grad_norm": 0.005349478218704462, + "learning_rate": 1.993423839463052e-05, + "loss": 9.109645179705694e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00009, + "step": 408, + "tokens/total": 12345904, + "tokens/train_per_sec_per_gpu": 32.78, + "tokens/trainable": 186659 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011329016648232937, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00016253325156867504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12376336, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 187083 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.019948504865169525, + "learning_rate": 1.957330080137385e-05, + "loss": 0.00013768397911917418, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 410, + "tokens/total": 12406880, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 187530 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.003809648798778653, + "learning_rate": 1.9395039639322864e-05, + "loss": 5.0881026254501194e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 411, + "tokens/total": 12437088, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 187972 + }, + { + "epoch": 1.609375, + "grad_norm": 0.012643440626561642, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0001802633050829172, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00018, + "step": 412, + "tokens/total": 12465440, + "tokens/train_per_sec_per_gpu": 39.92, + "tokens/trainable": 188417 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.22178205847740173, + "learning_rate": 1.904296967493982e-05, + "loss": 0.011118180118501186, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01118, + "step": 413, + "tokens/total": 12495968, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 188868 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.0006713962648063898, + "learning_rate": 1.8869175523676064e-05, + "loss": 1.6992760720313527e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 414, + "tokens/total": 12526128, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 189337 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.006867737043648958, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00011090705083915964, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00011, + "step": 415, + "tokens/total": 12556256, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 189750 + }, + { + "epoch": 1.625, + "grad_norm": 0.016067150980234146, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0001882174692582339, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 416, + "tokens/total": 12586736, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 190216 + } + ], + "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": 8.543353259829796e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-416/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-448/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-448/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-448/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..39dced0968756f01a03b1a595752ea77473a5679 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78745c3cd6de0abeb5c17fb39d616fa0b058724c684b4b52e35e350df066281e +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-448/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..6aacccb97703af5f66af810640b3823e9e9c2f41 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bde78bc929d23d7ca8c2c487e8765f625e9e138fee7bb24ac85a1ff732ee721 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..a848c2b6c5f74a7dfcb7aeaf75adc3de783f7394 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0f5fea350892f85d3bbab116a9d5db62bdcedcd47807fd37fd8d790fcd8718a +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2045249a1135f9f2d87c87c3e02e6227697d0cfa --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-448/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-448/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-448/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c6796d0904681cc1c501bb821a45ca8b7860aa52 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/tokens_state.json @@ -0,0 +1 @@ +{"total": 13557392, "trainable": 204830} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3e2039d0dbe543c540a5baca9f1803d7bd3bacca --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.03220707178115845, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.000319857761496678, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00032, + "step": 289, + "tokens/total": 8739680, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 132397 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.031606193631887436, + "learning_rate": 4.90663667927174e-05, + "loss": 0.0005622455500997603, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00056, + "step": 290, + "tokens/total": 8770112, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 132846 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.6305046677589417, + "learning_rate": 4.877873600900581e-05, + "loss": 0.006522824056446552, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00654, + "step": 291, + "tokens/total": 8800368, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 133347 + }, + { + "epoch": 1.140625, + "grad_norm": 0.03814006596803665, + "learning_rate": 4.849136411748306e-05, + "loss": 0.000703076075296849, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0007, + "step": 292, + "tokens/total": 8830688, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 133779 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.05072004348039627, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0011388614075258374, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00114, + "step": 293, + "tokens/total": 8861008, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 134251 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.2358824610710144, + "learning_rate": 4.791744483460251e-05, + "loss": 0.006398226134479046, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00642, + "step": 294, + "tokens/total": 8891760, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 134683 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2783265709877014, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0070259119383990765, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00705, + "step": 295, + "tokens/total": 8922192, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 135165 + }, + { + "epoch": 1.15625, + "grad_norm": 0.34820756316185, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.004574563354253769, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 296, + "tokens/total": 8952496, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 135599 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.03794313222169876, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0009930897504091263, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 297, + "tokens/total": 8982768, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 136038 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.19601035118103027, + "learning_rate": 4.677323837567412e-05, + "loss": 0.003028921550139785, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00303, + "step": 298, + "tokens/total": 9013248, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 136495 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0572715662419796, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.001314631663262844, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00132, + "step": 299, + "tokens/total": 9043632, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 136953 + }, + { + "epoch": 1.171875, + "grad_norm": 0.12685413658618927, + "learning_rate": 4.620314165804964e-05, + "loss": 0.0023695232812315226, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00237, + "step": 300, + "tokens/total": 9073936, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 137412 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.04133208841085434, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0008228939259424806, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00082, + "step": 301, + "tokens/total": 9104416, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 137907 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.01651018112897873, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0003794343792833388, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00038, + "step": 302, + "tokens/total": 9134688, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 138383 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.009257322177290916, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00024720095098018646, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00025, + "step": 303, + "tokens/total": 9165088, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 138842 + }, + { + "epoch": 1.1875, + "grad_norm": 0.038815200328826904, + "learning_rate": 4.506743570677743e-05, + "loss": 0.000565327238291502, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 304, + "tokens/total": 9195584, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 139269 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.37655678391456604, + "learning_rate": 4.478451305763618e-05, + "loss": 0.005295810755342245, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00531, + "step": 305, + "tokens/total": 9225760, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 139736 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.014935888350009918, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0003102960763499141, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00031, + "step": 306, + "tokens/total": 9256352, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 140226 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.02618814818561077, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0003336220979690552, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00033, + "step": 307, + "tokens/total": 9284256, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 140664 + }, + { + "epoch": 1.203125, + "grad_norm": 0.2529093027114868, + "learning_rate": 4.393834276419352e-05, + "loss": 0.005605834536254406, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00562, + "step": 308, + "tokens/total": 9314768, + "tokens/train_per_sec_per_gpu": 29.95, + "tokens/trainable": 141115 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0079645412042737, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00018522725440561771, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 309, + "tokens/total": 9344976, + "tokens/train_per_sec_per_gpu": 28.95, + "tokens/trainable": 141530 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.4591776132583618, + "learning_rate": 4.337651127546448e-05, + "loss": 0.003637854475528002, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00364, + "step": 310, + "tokens/total": 9375280, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 141964 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.02764446660876274, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00038255995605140924, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00038, + "step": 311, + "tokens/total": 9405536, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 142428 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0073210448026657104, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.0001625180011615157, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 312, + "tokens/total": 9435696, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 142894 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.1570003777742386, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00023247752687893808, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00023, + "step": 313, + "tokens/total": 9466080, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 143368 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.2202703058719635, + "learning_rate": 4.225874584473174e-05, + "loss": 0.002779184840619564, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00278, + "step": 314, + "tokens/total": 9496400, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 143824 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.04489271715283394, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0008375659235753119, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00084, + "step": 315, + "tokens/total": 9526880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 144306 + }, + { + "epoch": 1.234375, + "grad_norm": 0.006173362489789724, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001250750501640141, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 316, + "tokens/total": 9557280, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 144763 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.011352705769240856, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00022372312378138304, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 317, + "tokens/total": 9587712, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 145255 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.05640754476189613, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0007214375073090196, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00072, + "step": 318, + "tokens/total": 9618064, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 145714 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.15869443118572235, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0053039658814668655, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00532, + "step": 319, + "tokens/total": 9648336, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 146200 + }, + { + "epoch": 1.25, + "grad_norm": 0.44812121987342834, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0036084880121052265, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00362, + "step": 320, + "tokens/total": 9678688, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 146668 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.05665228143334389, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0009465725161135197, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00095, + "step": 321, + "tokens/total": 9708848, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 147114 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.016061583533883095, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00015383574645966291, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00015, + "step": 322, + "tokens/total": 9739360, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 147564 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.06347034126520157, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0009517542785033584, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00095, + "step": 323, + "tokens/total": 9769776, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 148013 + }, + { + "epoch": 1.265625, + "grad_norm": 0.24154019355773926, + "learning_rate": 3.95030593412612e-05, + "loss": 0.004087492823600769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.0041, + "step": 324, + "tokens/total": 9800096, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 148470 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.22498610615730286, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0027766539715230465, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00278, + "step": 325, + "tokens/total": 9830304, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 148926 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.002810331992805004, + "learning_rate": 3.895929566172861e-05, + "loss": 4.6118868340272456e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 326, + "tokens/total": 9860608, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 149420 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.007315841969102621, + "learning_rate": 3.868840945050728e-05, + "loss": 0.00011396597255952656, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00011, + "step": 327, + "tokens/total": 9890560, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 149873 + }, + { + "epoch": 1.28125, + "grad_norm": 0.021338002756237984, + "learning_rate": 3.841820203114995e-05, + "loss": 0.00021865982853341848, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 328, + "tokens/total": 9920880, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 150347 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.027939820662140846, + "learning_rate": 3.814868464809027e-05, + "loss": 0.0003473583492450416, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 329, + "tokens/total": 9951280, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 150822 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.0030000312253832817, + "learning_rate": 3.787986851704667e-05, + "loss": 5.043415876571089e-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": 9981600, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 151266 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.016975272446870804, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.0002860842505469918, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 331, + "tokens/total": 10012016, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 151734 + }, + { + "epoch": 1.296875, + "grad_norm": 0.0012887469492852688, + "learning_rate": 3.734438472750619e-05, + "loss": 2.801024675136432e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 332, + "tokens/total": 10042448, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 152184 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.0035721107851713896, + "learning_rate": 3.707773935267552e-05, + "loss": 4.7658577386755496e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 333, + "tokens/total": 10073024, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 152623 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.036152660846710205, + "learning_rate": 3.68118397962661e-05, + "loss": 0.00040513710700906813, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00041, + "step": 334, + "tokens/total": 10103504, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 153040 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.11260190606117249, + "learning_rate": 3.654669712344384e-05, + "loss": 0.0015010370407253504, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.0015, + "step": 335, + "tokens/total": 10131632, + "tokens/train_per_sec_per_gpu": 31.76, + "tokens/trainable": 153463 + }, + { + "epoch": 1.3125, + "grad_norm": 0.17602311074733734, + "learning_rate": 3.628232236787763e-05, + "loss": 0.007916431874036789, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00795, + "step": 336, + "tokens/total": 10161824, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 153908 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.005530365742743015, + "learning_rate": 3.6018726531280144e-05, + "loss": 8.388348942389712e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00008, + "step": 337, + "tokens/total": 10192448, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 154387 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.06029028072953224, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0002887388109229505, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00029, + "step": 338, + "tokens/total": 10222704, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 154836 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.003610784187912941, + "learning_rate": 3.549391545931585e-05, + "loss": 6.314883648883551e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 339, + "tokens/total": 10253168, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 155261 + }, + { + "epoch": 1.328125, + "grad_norm": 0.0029664942994713783, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.0910730604082346e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10283552, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 155674 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.3301459848880768, + "learning_rate": 3.49723512647657e-05, + "loss": 0.001420877524651587, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 341, + "tokens/total": 10313872, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 156164 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.0065552121959626675, + "learning_rate": 3.471281389826491e-05, + "loss": 3.5097506042802706e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00004, + "step": 342, + "tokens/total": 10344256, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 156611 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.027132108807563782, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.00014500679390039295, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00015, + "step": 343, + "tokens/total": 10374544, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 157073 + }, + { + "epoch": 1.34375, + "grad_norm": 0.40091755986213684, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.004070552997291088, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00408, + "step": 344, + "tokens/total": 10405040, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 157523 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.2770630419254303, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.005590756889432669, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00561, + "step": 345, + "tokens/total": 10435424, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 157964 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.007484348025172949, + "learning_rate": 3.3683214232914404e-05, + "loss": 8.245774370152503e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 346, + "tokens/total": 10465760, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 158390 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005284447688609362, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00010297571134287864, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 347, + "tokens/total": 10495904, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 158848 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0784892812371254, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0006088463123887777, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00061, + "step": 348, + "tokens/total": 10526624, + "tokens/train_per_sec_per_gpu": 28.7, + "tokens/trainable": 159270 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.33535251021385193, + "learning_rate": 3.292029118616024e-05, + "loss": 0.006133331451565027, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00615, + "step": 349, + "tokens/total": 10556752, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 159720 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.008170238696038723, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00012147967208875343, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00012, + "step": 350, + "tokens/total": 10587264, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 160187 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.001569622429087758, + "learning_rate": 3.241625231705354e-05, + "loss": 3.8519159716088325e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 351, + "tokens/total": 10617504, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 160645 + }, + { + "epoch": 1.375, + "grad_norm": 0.143654003739357, + "learning_rate": 3.216563735127618e-05, + "loss": 0.002758543938398361, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00276, + "step": 352, + "tokens/total": 10647760, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 161102 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.042433012276887894, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00035650472273118794, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 353, + "tokens/total": 10678080, + "tokens/train_per_sec_per_gpu": 28.87, + "tokens/trainable": 161555 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.9453999996185303, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0030506744515150785, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00306, + "step": 354, + "tokens/total": 10708544, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 162020 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.03272778168320656, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00023931957548484206, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00024, + "step": 355, + "tokens/total": 10739024, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 162448 + }, + { + "epoch": 1.390625, + "grad_norm": 0.006854584906250238, + "learning_rate": 3.11727834939056e-05, + "loss": 6.610051059396937e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10769024, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 162863 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.03258929029107094, + "learning_rate": 3.092702317708967e-05, + "loss": 0.0002595578844193369, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00026, + "step": 357, + "tokens/total": 10799328, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 163318 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.21498654782772064, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.003778135636821389, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00379, + "step": 358, + "tokens/total": 10829488, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 163767 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.22063465416431427, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.00428510969504714, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00429, + "step": 359, + "tokens/total": 10859936, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 164208 + }, + { + "epoch": 1.40625, + "grad_norm": 0.01306198537349701, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00016752032388467342, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00017, + "step": 360, + "tokens/total": 10890400, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 164655 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.42359107732772827, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.022302038967609406, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02255, + "step": 361, + "tokens/total": 10920624, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 165100 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.14819112420082092, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.002135487971827388, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00214, + "step": 362, + "tokens/total": 10951136, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 165587 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.10172182321548462, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.0012491067172959447, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00125, + "step": 363, + "tokens/total": 10981344, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 166034 + }, + { + "epoch": 1.421875, + "grad_norm": 0.16598454117774963, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0014915807405486703, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00149, + "step": 364, + "tokens/total": 11011552, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 166485 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.01892480067908764, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00023381829669233412, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00023, + "step": 365, + "tokens/total": 11041872, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 166973 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.050091467797756195, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00035884848330169916, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00036, + "step": 366, + "tokens/total": 11072192, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 167414 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.6703732013702393, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.008329304866492748, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00836, + "step": 367, + "tokens/total": 11102736, + "tokens/train_per_sec_per_gpu": 30.18, + "tokens/trainable": 167851 + }, + { + "epoch": 1.4375, + "grad_norm": 0.10087089240550995, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0014869305305182934, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00149, + "step": 368, + "tokens/total": 11133056, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 168294 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.03148820623755455, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00043267637374810874, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 369, + "tokens/total": 11163568, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 168769 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.01531990710645914, + "learning_rate": 2.782696506053033e-05, + "loss": 0.00021343299886211753, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00021, + "step": 370, + "tokens/total": 11193936, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 169270 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.006444540806114674, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.00014627019118051976, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00015, + "step": 371, + "tokens/total": 11224480, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 169711 + }, + { + "epoch": 1.453125, + "grad_norm": 0.31427323818206787, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.00462032575160265, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00463, + "step": 372, + "tokens/total": 11254912, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 170185 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.013480601832270622, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.000224081173655577, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00022, + "step": 373, + "tokens/total": 11285104, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 170650 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.023016467690467834, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0001827064115786925, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00018, + "step": 374, + "tokens/total": 11315600, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 171126 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.019638920202851295, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0005156319821253419, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00052, + "step": 375, + "tokens/total": 11345776, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 171600 + }, + { + "epoch": 1.46875, + "grad_norm": 0.03251553326845169, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004913917509838939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00049, + "step": 376, + "tokens/total": 11376208, + "tokens/train_per_sec_per_gpu": 28.1, + "tokens/trainable": 172017 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.03556421399116516, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0005220613675191998, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00052, + "step": 377, + "tokens/total": 11406512, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 172493 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.04202103987336159, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0004697944095823914, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00047, + "step": 378, + "tokens/total": 11436544, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 172951 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.023767679929733276, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0004130054439883679, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00041, + "step": 379, + "tokens/total": 11467008, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 173435 + }, + { + "epoch": 1.484375, + "grad_norm": 0.04106076434254646, + "learning_rate": 2.557117581955798e-05, + "loss": 0.00047382546472363174, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00047, + "step": 380, + "tokens/total": 11497184, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 173900 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.15864700078964233, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.003480169689282775, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00349, + "step": 381, + "tokens/total": 11527600, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 174371 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.04146522656083107, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0005609758081845939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00056, + "step": 382, + "tokens/total": 11557808, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 174826 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.012171853333711624, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00022788115893490613, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00023, + "step": 383, + "tokens/total": 11588352, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 175276 + }, + { + "epoch": 1.5, + "grad_norm": 0.08181510865688324, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0016442432533949614, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 384, + "tokens/total": 11618640, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 175720 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.10246019810438156, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0017187006305903196, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00172, + "step": 385, + "tokens/total": 11649072, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 176203 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.2453320175409317, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0038096809294074774, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00382, + "step": 386, + "tokens/total": 11679552, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 176651 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.005247004795819521, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00010425677464809269, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 387, + "tokens/total": 11709760, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 177115 + }, + { + "epoch": 1.515625, + "grad_norm": 0.027557479217648506, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00036113697569817305, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00036, + "step": 388, + "tokens/total": 11740288, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 177601 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.028131073340773582, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0005047487211413682, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0005, + "step": 389, + "tokens/total": 11770592, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 178050 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.0033069071359932423, + "learning_rate": 2.3437809889624914e-05, + "loss": 7.658063259441406e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00008, + "step": 390, + "tokens/total": 11800752, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 178531 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.028486015275120735, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0005519004189409316, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00055, + "step": 391, + "tokens/total": 11831360, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 179000 + }, + { + "epoch": 1.53125, + "grad_norm": 0.002584393136203289, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.4768250265624374e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 392, + "tokens/total": 11861552, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 179457 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.009344775229692459, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.0002210603270214051, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00022, + "step": 393, + "tokens/total": 11891904, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 179890 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.004884887952357531, + "learning_rate": 2.2620747959533722e-05, + "loss": 8.428651199210435e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00008, + "step": 394, + "tokens/total": 11922208, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 180341 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.023302890360355377, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.000332253344822675, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00033, + "step": 395, + "tokens/total": 11952672, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 180762 + }, + { + "epoch": 1.546875, + "grad_norm": 0.14256854355335236, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00286297220736742, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00287, + "step": 396, + "tokens/total": 11983088, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 181226 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.00555575592443347, + "learning_rate": 2.202206960760984e-05, + "loss": 9.771026816451922e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0001, + "step": 397, + "tokens/total": 12013488, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 181714 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.007696876302361488, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0001580697571625933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00016, + "step": 398, + "tokens/total": 12043968, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 182142 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.0037960875779390335, + "learning_rate": 2.1629798596457056e-05, + "loss": 7.467059913324192e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 399, + "tokens/total": 12074160, + "tokens/train_per_sec_per_gpu": 34.82, + "tokens/trainable": 182604 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004514187108725309, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.00011837383499369025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00012, + "step": 400, + "tokens/total": 12104352, + "tokens/train_per_sec_per_gpu": 25.26, + "tokens/trainable": 183011 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.09468130767345428, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0007745609618723392, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00077, + "step": 401, + "tokens/total": 12134240, + "tokens/train_per_sec_per_gpu": 31.75, + "tokens/trainable": 183459 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.017941495403647423, + "learning_rate": 2.105182715082638e-05, + "loss": 0.00024286848201882094, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 402, + "tokens/total": 12164480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 183959 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.005621996242552996, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.00012180282647022977, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00012, + "step": 403, + "tokens/total": 12194640, + "tokens/train_per_sec_per_gpu": 32.46, + "tokens/trainable": 184431 + }, + { + "epoch": 1.578125, + "grad_norm": 0.004712392576038837, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010195293725701049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 404, + "tokens/total": 12224960, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 184867 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.023734599351882935, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.00040538967004977167, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00041, + "step": 405, + "tokens/total": 12255008, + "tokens/train_per_sec_per_gpu": 29.33, + "tokens/trainable": 185300 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.06495791673660278, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.000537894491571933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00054, + "step": 406, + "tokens/total": 12285248, + "tokens/train_per_sec_per_gpu": 31.47, + "tokens/trainable": 185754 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.003490478964522481, + "learning_rate": 2.011689980574966e-05, + "loss": 6.170629058033228e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00006, + "step": 407, + "tokens/total": 12315552, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 186213 + }, + { + "epoch": 1.59375, + "grad_norm": 0.005349478218704462, + "learning_rate": 1.993423839463052e-05, + "loss": 9.109645179705694e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00009, + "step": 408, + "tokens/total": 12345904, + "tokens/train_per_sec_per_gpu": 32.78, + "tokens/trainable": 186659 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011329016648232937, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00016253325156867504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12376336, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 187083 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.019948504865169525, + "learning_rate": 1.957330080137385e-05, + "loss": 0.00013768397911917418, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 410, + "tokens/total": 12406880, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 187530 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.003809648798778653, + "learning_rate": 1.9395039639322864e-05, + "loss": 5.0881026254501194e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 411, + "tokens/total": 12437088, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 187972 + }, + { + "epoch": 1.609375, + "grad_norm": 0.012643440626561642, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0001802633050829172, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00018, + "step": 412, + "tokens/total": 12465440, + "tokens/train_per_sec_per_gpu": 39.92, + "tokens/trainable": 188417 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.22178205847740173, + "learning_rate": 1.904296967493982e-05, + "loss": 0.011118180118501186, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01118, + "step": 413, + "tokens/total": 12495968, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 188868 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.0006713962648063898, + "learning_rate": 1.8869175523676064e-05, + "loss": 1.6992760720313527e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 414, + "tokens/total": 12526128, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 189337 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.006867737043648958, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00011090705083915964, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00011, + "step": 415, + "tokens/total": 12556256, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 189750 + }, + { + "epoch": 1.625, + "grad_norm": 0.016067150980234146, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0001882174692582339, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 416, + "tokens/total": 12586736, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 190216 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0036702004726976156, + "learning_rate": 1.8356842992397304e-05, + "loss": 4.983436519978568e-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": 12617168, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 190708 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.3933418095111847, + "learning_rate": 1.8189105812005714e-05, + "loss": 0.0050767892971634865, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00509, + "step": 418, + "tokens/total": 12647680, + "tokens/train_per_sec_per_gpu": 30.59, + "tokens/trainable": 191126 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.04893051087856293, + "learning_rate": 1.802290048317732e-05, + "loss": 0.0004771970270667225, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00048, + "step": 419, + "tokens/total": 12677952, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 191540 + }, + { + "epoch": 1.640625, + "grad_norm": 0.03733966872096062, + "learning_rate": 1.785823392239424e-05, + "loss": 0.0004862284695263952, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00049, + "step": 420, + "tokens/total": 12708416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.00783330574631691, + "learning_rate": 1.7695112982104225e-05, + "loss": 7.455339073203504e-05, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 421, + "tokens/total": 12738640, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 192463 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.052442099899053574, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0009662319207563996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00097, + "step": 422, + "tokens/total": 12769232, + "tokens/train_per_sec_per_gpu": 32.64, + "tokens/trainable": 192918 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.024694403633475304, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.00027854571817442775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00028, + "step": 423, + "tokens/total": 12799648, + "tokens/train_per_sec_per_gpu": 37.36, + "tokens/trainable": 193425 + }, + { + "epoch": 1.65625, + "grad_norm": 0.002282459754496813, + "learning_rate": 1.721509144218405e-05, + "loss": 5.6570435845060274e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 424, + "tokens/total": 12829920, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 193898 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.28251373767852783, + "learning_rate": 1.705822021773101e-05, + "loss": 0.010694734752178192, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01075, + "step": 425, + "tokens/total": 12860112, + "tokens/train_per_sec_per_gpu": 29.22, + "tokens/trainable": 194315 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.14038732647895813, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0015529554802924395, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00155, + "step": 426, + "tokens/total": 12890320, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 194796 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.0007188008166849613, + "learning_rate": 1.6749220968158415e-05, + "loss": 1.8845212252927013e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00002, + "step": 427, + "tokens/total": 12920656, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 195262 + }, + { + "epoch": 1.671875, + "grad_norm": 0.002983702579513192, + "learning_rate": 1.659710580175893e-05, + "loss": 5.061700358055532e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 428, + "tokens/total": 12951040, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 195680 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.00501316599547863, + "learning_rate": 1.644658873654133e-05, + "loss": 8.430491288891062e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00008, + "step": 429, + "tokens/total": 12981232, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 196145 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.6195209622383118, + "learning_rate": 1.629767603613508e-05, + "loss": 0.0007395793218165636, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00074, + "step": 430, + "tokens/total": 13011616, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 196626 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.32764554023742676, + "learning_rate": 1.615037389740547e-05, + "loss": 0.012379031628370285, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.01246, + "step": 431, + "tokens/total": 13042208, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 197067 + }, + { + "epoch": 1.6875, + "grad_norm": 0.28968432545661926, + "learning_rate": 1.600468845019576e-05, + "loss": 0.003729558549821377, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00374, + "step": 432, + "tokens/total": 13072800, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 197565 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.001470765215344727, + "learning_rate": 1.5860625757072092e-05, + "loss": 3.539394310791977e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00004, + "step": 433, + "tokens/total": 13103328, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 198015 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.019142981618642807, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00024474930251017213, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 434, + "tokens/total": 13133472, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 198442 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.009595178067684174, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0001564246922498569, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00016, + "step": 435, + "tokens/total": 13164064, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 198914 + }, + { + "epoch": 1.703125, + "grad_norm": 0.22099021077156067, + "learning_rate": 1.543823381344311e-05, + "loss": 0.014420191757380962, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01452, + "step": 436, + "tokens/total": 13194464, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 199355 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.1727798581123352, + "learning_rate": 1.5300721408011114e-05, + "loss": 0.008166543208062649, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0082, + "step": 437, + "tokens/total": 13224992, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 199830 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.015094809234142303, + "learning_rate": 1.5164861051607254e-05, + "loss": 0.0002187314530601725, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 438, + "tokens/total": 13255296, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 200338 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.07529865950345993, + "learning_rate": 1.5030658397935521e-05, + "loss": 0.0006286898860707879, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00063, + "step": 439, + "tokens/total": 13285344, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 200785 + }, + { + "epoch": 1.71875, + "grad_norm": 0.005869010463356972, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00010347182978875935, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 440, + "tokens/total": 13315632, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 201267 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.02751794271171093, + "learning_rate": 1.476724846845306e-05, + "loss": 0.000567559851333499, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00057, + "step": 441, + "tokens/total": 13346048, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 201734 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.06305643171072006, + "learning_rate": 1.463805215420471e-05, + "loss": 0.0012271346058696508, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00123, + "step": 442, + "tokens/total": 13376304, + "tokens/train_per_sec_per_gpu": 30.3, + "tokens/trainable": 202174 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.008338473737239838, + "learning_rate": 1.451053546535705e-05, + "loss": 0.00016968630370683968, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00017, + "step": 443, + "tokens/total": 13406720, + "tokens/train_per_sec_per_gpu": 36.21, + "tokens/trainable": 202612 + }, + { + "epoch": 1.734375, + "grad_norm": 0.020954634994268417, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0005087319877929986, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00051, + "step": 444, + "tokens/total": 13436464, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 203020 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.019224021583795547, + "learning_rate": 1.4260562119706606e-05, + "loss": 0.0003236978664062917, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00032, + "step": 445, + "tokens/total": 13466672, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 203479 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.02892594039440155, + "learning_rate": 1.413811586531508e-05, + "loss": 0.0004480727366171777, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00045, + "step": 446, + "tokens/total": 13496736, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 203946 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.013984446413815022, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.00018316772184334695, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00018, + "step": 447, + "tokens/total": 13526928, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 204382 + }, + { + "epoch": 1.75, + "grad_norm": 0.009344048798084259, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.00011864845873787999, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00012, + "step": 448, + "tokens/total": 13557392, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 204830 + } + ], + "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": 9.202194209681556e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-448/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-480/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-480/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-480/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..fe4b00b457dc0e42ac0b5f42cf07a3d8863411fa --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:67a90c20a33d71b9244dde6465f0a418fa3029218d6c45c220470ea5e475440e +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-480/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..d69aaf5e4472df76e5d996900b2c87c9a4752993 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2869c0e034133e5665035d67617933a04a65a75601c2b2d8efe51f7f27fb6a9 +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..94daead8b73acb85d313bc316b68b7b7c8ba7b20 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6cc653826760fd60176ec4de753ff8ff573e43d06826f127f29d6db37ba7f969 +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce1c4e63a6c6f8be8c137d3add4eda184d59a043 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-480/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-480/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-480/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..5bb8973ac2042035241f89bc5e72ffe120b40fbb --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/tokens_state.json @@ -0,0 +1 @@ +{"total": 14522544, "trainable": 219350} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a2b0b4b3504bb61be9782537be3ab19246fbe4eb --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.03220707178115845, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.000319857761496678, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00032, + "step": 289, + "tokens/total": 8739680, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 132397 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.031606193631887436, + "learning_rate": 4.90663667927174e-05, + "loss": 0.0005622455500997603, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00056, + "step": 290, + "tokens/total": 8770112, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 132846 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.6305046677589417, + "learning_rate": 4.877873600900581e-05, + "loss": 0.006522824056446552, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00654, + "step": 291, + "tokens/total": 8800368, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 133347 + }, + { + "epoch": 1.140625, + "grad_norm": 0.03814006596803665, + "learning_rate": 4.849136411748306e-05, + "loss": 0.000703076075296849, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0007, + "step": 292, + "tokens/total": 8830688, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 133779 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.05072004348039627, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0011388614075258374, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00114, + "step": 293, + "tokens/total": 8861008, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 134251 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.2358824610710144, + "learning_rate": 4.791744483460251e-05, + "loss": 0.006398226134479046, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00642, + "step": 294, + "tokens/total": 8891760, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 134683 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2783265709877014, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0070259119383990765, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00705, + "step": 295, + "tokens/total": 8922192, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 135165 + }, + { + "epoch": 1.15625, + "grad_norm": 0.34820756316185, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.004574563354253769, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 296, + "tokens/total": 8952496, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 135599 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.03794313222169876, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0009930897504091263, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 297, + "tokens/total": 8982768, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 136038 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.19601035118103027, + "learning_rate": 4.677323837567412e-05, + "loss": 0.003028921550139785, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00303, + "step": 298, + "tokens/total": 9013248, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 136495 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0572715662419796, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.001314631663262844, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00132, + "step": 299, + "tokens/total": 9043632, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 136953 + }, + { + "epoch": 1.171875, + "grad_norm": 0.12685413658618927, + "learning_rate": 4.620314165804964e-05, + "loss": 0.0023695232812315226, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00237, + "step": 300, + "tokens/total": 9073936, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 137412 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.04133208841085434, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0008228939259424806, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00082, + "step": 301, + "tokens/total": 9104416, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 137907 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.01651018112897873, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0003794343792833388, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00038, + "step": 302, + "tokens/total": 9134688, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 138383 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.009257322177290916, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00024720095098018646, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00025, + "step": 303, + "tokens/total": 9165088, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 138842 + }, + { + "epoch": 1.1875, + "grad_norm": 0.038815200328826904, + "learning_rate": 4.506743570677743e-05, + "loss": 0.000565327238291502, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 304, + "tokens/total": 9195584, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 139269 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.37655678391456604, + "learning_rate": 4.478451305763618e-05, + "loss": 0.005295810755342245, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00531, + "step": 305, + "tokens/total": 9225760, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 139736 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.014935888350009918, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0003102960763499141, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00031, + "step": 306, + "tokens/total": 9256352, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 140226 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.02618814818561077, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0003336220979690552, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00033, + "step": 307, + "tokens/total": 9284256, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 140664 + }, + { + "epoch": 1.203125, + "grad_norm": 0.2529093027114868, + "learning_rate": 4.393834276419352e-05, + "loss": 0.005605834536254406, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00562, + "step": 308, + "tokens/total": 9314768, + "tokens/train_per_sec_per_gpu": 29.95, + "tokens/trainable": 141115 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0079645412042737, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00018522725440561771, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 309, + "tokens/total": 9344976, + "tokens/train_per_sec_per_gpu": 28.95, + "tokens/trainable": 141530 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.4591776132583618, + "learning_rate": 4.337651127546448e-05, + "loss": 0.003637854475528002, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00364, + "step": 310, + "tokens/total": 9375280, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 141964 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.02764446660876274, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00038255995605140924, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00038, + "step": 311, + "tokens/total": 9405536, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 142428 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0073210448026657104, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.0001625180011615157, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 312, + "tokens/total": 9435696, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 142894 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.1570003777742386, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00023247752687893808, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00023, + "step": 313, + "tokens/total": 9466080, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 143368 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.2202703058719635, + "learning_rate": 4.225874584473174e-05, + "loss": 0.002779184840619564, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00278, + "step": 314, + "tokens/total": 9496400, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 143824 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.04489271715283394, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0008375659235753119, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00084, + "step": 315, + "tokens/total": 9526880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 144306 + }, + { + "epoch": 1.234375, + "grad_norm": 0.006173362489789724, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001250750501640141, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 316, + "tokens/total": 9557280, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 144763 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.011352705769240856, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00022372312378138304, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 317, + "tokens/total": 9587712, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 145255 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.05640754476189613, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0007214375073090196, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00072, + "step": 318, + "tokens/total": 9618064, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 145714 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.15869443118572235, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0053039658814668655, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00532, + "step": 319, + "tokens/total": 9648336, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 146200 + }, + { + "epoch": 1.25, + "grad_norm": 0.44812121987342834, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0036084880121052265, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00362, + "step": 320, + "tokens/total": 9678688, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 146668 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.05665228143334389, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0009465725161135197, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00095, + "step": 321, + "tokens/total": 9708848, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 147114 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.016061583533883095, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00015383574645966291, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00015, + "step": 322, + "tokens/total": 9739360, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 147564 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.06347034126520157, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0009517542785033584, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00095, + "step": 323, + "tokens/total": 9769776, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 148013 + }, + { + "epoch": 1.265625, + "grad_norm": 0.24154019355773926, + "learning_rate": 3.95030593412612e-05, + "loss": 0.004087492823600769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.0041, + "step": 324, + "tokens/total": 9800096, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 148470 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.22498610615730286, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0027766539715230465, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00278, + "step": 325, + "tokens/total": 9830304, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 148926 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.002810331992805004, + "learning_rate": 3.895929566172861e-05, + "loss": 4.6118868340272456e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 326, + "tokens/total": 9860608, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 149420 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.007315841969102621, + "learning_rate": 3.868840945050728e-05, + "loss": 0.00011396597255952656, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00011, + "step": 327, + "tokens/total": 9890560, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 149873 + }, + { + "epoch": 1.28125, + "grad_norm": 0.021338002756237984, + "learning_rate": 3.841820203114995e-05, + "loss": 0.00021865982853341848, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 328, + "tokens/total": 9920880, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 150347 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.027939820662140846, + "learning_rate": 3.814868464809027e-05, + "loss": 0.0003473583492450416, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 329, + "tokens/total": 9951280, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 150822 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.0030000312253832817, + "learning_rate": 3.787986851704667e-05, + "loss": 5.043415876571089e-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": 9981600, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 151266 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.016975272446870804, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.0002860842505469918, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 331, + "tokens/total": 10012016, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 151734 + }, + { + "epoch": 1.296875, + "grad_norm": 0.0012887469492852688, + "learning_rate": 3.734438472750619e-05, + "loss": 2.801024675136432e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 332, + "tokens/total": 10042448, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 152184 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.0035721107851713896, + "learning_rate": 3.707773935267552e-05, + "loss": 4.7658577386755496e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 333, + "tokens/total": 10073024, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 152623 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.036152660846710205, + "learning_rate": 3.68118397962661e-05, + "loss": 0.00040513710700906813, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00041, + "step": 334, + "tokens/total": 10103504, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 153040 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.11260190606117249, + "learning_rate": 3.654669712344384e-05, + "loss": 0.0015010370407253504, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.0015, + "step": 335, + "tokens/total": 10131632, + "tokens/train_per_sec_per_gpu": 31.76, + "tokens/trainable": 153463 + }, + { + "epoch": 1.3125, + "grad_norm": 0.17602311074733734, + "learning_rate": 3.628232236787763e-05, + "loss": 0.007916431874036789, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00795, + "step": 336, + "tokens/total": 10161824, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 153908 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.005530365742743015, + "learning_rate": 3.6018726531280144e-05, + "loss": 8.388348942389712e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00008, + "step": 337, + "tokens/total": 10192448, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 154387 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.06029028072953224, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0002887388109229505, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00029, + "step": 338, + "tokens/total": 10222704, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 154836 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.003610784187912941, + "learning_rate": 3.549391545931585e-05, + "loss": 6.314883648883551e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 339, + "tokens/total": 10253168, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 155261 + }, + { + "epoch": 1.328125, + "grad_norm": 0.0029664942994713783, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.0910730604082346e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10283552, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 155674 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.3301459848880768, + "learning_rate": 3.49723512647657e-05, + "loss": 0.001420877524651587, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 341, + "tokens/total": 10313872, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 156164 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.0065552121959626675, + "learning_rate": 3.471281389826491e-05, + "loss": 3.5097506042802706e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00004, + "step": 342, + "tokens/total": 10344256, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 156611 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.027132108807563782, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.00014500679390039295, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00015, + "step": 343, + "tokens/total": 10374544, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 157073 + }, + { + "epoch": 1.34375, + "grad_norm": 0.40091755986213684, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.004070552997291088, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00408, + "step": 344, + "tokens/total": 10405040, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 157523 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.2770630419254303, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.005590756889432669, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00561, + "step": 345, + "tokens/total": 10435424, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 157964 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.007484348025172949, + "learning_rate": 3.3683214232914404e-05, + "loss": 8.245774370152503e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 346, + "tokens/total": 10465760, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 158390 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005284447688609362, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00010297571134287864, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 347, + "tokens/total": 10495904, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 158848 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0784892812371254, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0006088463123887777, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00061, + "step": 348, + "tokens/total": 10526624, + "tokens/train_per_sec_per_gpu": 28.7, + "tokens/trainable": 159270 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.33535251021385193, + "learning_rate": 3.292029118616024e-05, + "loss": 0.006133331451565027, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00615, + "step": 349, + "tokens/total": 10556752, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 159720 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.008170238696038723, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00012147967208875343, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00012, + "step": 350, + "tokens/total": 10587264, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 160187 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.001569622429087758, + "learning_rate": 3.241625231705354e-05, + "loss": 3.8519159716088325e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 351, + "tokens/total": 10617504, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 160645 + }, + { + "epoch": 1.375, + "grad_norm": 0.143654003739357, + "learning_rate": 3.216563735127618e-05, + "loss": 0.002758543938398361, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00276, + "step": 352, + "tokens/total": 10647760, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 161102 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.042433012276887894, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00035650472273118794, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 353, + "tokens/total": 10678080, + "tokens/train_per_sec_per_gpu": 28.87, + "tokens/trainable": 161555 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.9453999996185303, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0030506744515150785, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00306, + "step": 354, + "tokens/total": 10708544, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 162020 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.03272778168320656, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00023931957548484206, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00024, + "step": 355, + "tokens/total": 10739024, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 162448 + }, + { + "epoch": 1.390625, + "grad_norm": 0.006854584906250238, + "learning_rate": 3.11727834939056e-05, + "loss": 6.610051059396937e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10769024, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 162863 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.03258929029107094, + "learning_rate": 3.092702317708967e-05, + "loss": 0.0002595578844193369, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00026, + "step": 357, + "tokens/total": 10799328, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 163318 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.21498654782772064, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.003778135636821389, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00379, + "step": 358, + "tokens/total": 10829488, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 163767 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.22063465416431427, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.00428510969504714, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00429, + "step": 359, + "tokens/total": 10859936, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 164208 + }, + { + "epoch": 1.40625, + "grad_norm": 0.01306198537349701, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00016752032388467342, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00017, + "step": 360, + "tokens/total": 10890400, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 164655 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.42359107732772827, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.022302038967609406, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02255, + "step": 361, + "tokens/total": 10920624, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 165100 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.14819112420082092, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.002135487971827388, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00214, + "step": 362, + "tokens/total": 10951136, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 165587 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.10172182321548462, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.0012491067172959447, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00125, + "step": 363, + "tokens/total": 10981344, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 166034 + }, + { + "epoch": 1.421875, + "grad_norm": 0.16598454117774963, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0014915807405486703, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00149, + "step": 364, + "tokens/total": 11011552, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 166485 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.01892480067908764, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00023381829669233412, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00023, + "step": 365, + "tokens/total": 11041872, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 166973 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.050091467797756195, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00035884848330169916, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00036, + "step": 366, + "tokens/total": 11072192, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 167414 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.6703732013702393, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.008329304866492748, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00836, + "step": 367, + "tokens/total": 11102736, + "tokens/train_per_sec_per_gpu": 30.18, + "tokens/trainable": 167851 + }, + { + "epoch": 1.4375, + "grad_norm": 0.10087089240550995, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0014869305305182934, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00149, + "step": 368, + "tokens/total": 11133056, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 168294 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.03148820623755455, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00043267637374810874, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 369, + "tokens/total": 11163568, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 168769 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.01531990710645914, + "learning_rate": 2.782696506053033e-05, + "loss": 0.00021343299886211753, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00021, + "step": 370, + "tokens/total": 11193936, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 169270 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.006444540806114674, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.00014627019118051976, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00015, + "step": 371, + "tokens/total": 11224480, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 169711 + }, + { + "epoch": 1.453125, + "grad_norm": 0.31427323818206787, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.00462032575160265, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00463, + "step": 372, + "tokens/total": 11254912, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 170185 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.013480601832270622, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.000224081173655577, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00022, + "step": 373, + "tokens/total": 11285104, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 170650 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.023016467690467834, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0001827064115786925, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00018, + "step": 374, + "tokens/total": 11315600, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 171126 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.019638920202851295, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0005156319821253419, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00052, + "step": 375, + "tokens/total": 11345776, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 171600 + }, + { + "epoch": 1.46875, + "grad_norm": 0.03251553326845169, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004913917509838939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00049, + "step": 376, + "tokens/total": 11376208, + "tokens/train_per_sec_per_gpu": 28.1, + "tokens/trainable": 172017 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.03556421399116516, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0005220613675191998, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00052, + "step": 377, + "tokens/total": 11406512, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 172493 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.04202103987336159, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0004697944095823914, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00047, + "step": 378, + "tokens/total": 11436544, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 172951 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.023767679929733276, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0004130054439883679, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00041, + "step": 379, + "tokens/total": 11467008, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 173435 + }, + { + "epoch": 1.484375, + "grad_norm": 0.04106076434254646, + "learning_rate": 2.557117581955798e-05, + "loss": 0.00047382546472363174, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00047, + "step": 380, + "tokens/total": 11497184, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 173900 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.15864700078964233, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.003480169689282775, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00349, + "step": 381, + "tokens/total": 11527600, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 174371 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.04146522656083107, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0005609758081845939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00056, + "step": 382, + "tokens/total": 11557808, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 174826 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.012171853333711624, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00022788115893490613, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00023, + "step": 383, + "tokens/total": 11588352, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 175276 + }, + { + "epoch": 1.5, + "grad_norm": 0.08181510865688324, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0016442432533949614, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 384, + "tokens/total": 11618640, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 175720 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.10246019810438156, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0017187006305903196, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00172, + "step": 385, + "tokens/total": 11649072, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 176203 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.2453320175409317, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0038096809294074774, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00382, + "step": 386, + "tokens/total": 11679552, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 176651 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.005247004795819521, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00010425677464809269, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 387, + "tokens/total": 11709760, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 177115 + }, + { + "epoch": 1.515625, + "grad_norm": 0.027557479217648506, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00036113697569817305, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00036, + "step": 388, + "tokens/total": 11740288, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 177601 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.028131073340773582, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0005047487211413682, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0005, + "step": 389, + "tokens/total": 11770592, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 178050 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.0033069071359932423, + "learning_rate": 2.3437809889624914e-05, + "loss": 7.658063259441406e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00008, + "step": 390, + "tokens/total": 11800752, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 178531 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.028486015275120735, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0005519004189409316, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00055, + "step": 391, + "tokens/total": 11831360, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 179000 + }, + { + "epoch": 1.53125, + "grad_norm": 0.002584393136203289, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.4768250265624374e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 392, + "tokens/total": 11861552, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 179457 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.009344775229692459, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.0002210603270214051, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00022, + "step": 393, + "tokens/total": 11891904, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 179890 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.004884887952357531, + "learning_rate": 2.2620747959533722e-05, + "loss": 8.428651199210435e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00008, + "step": 394, + "tokens/total": 11922208, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 180341 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.023302890360355377, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.000332253344822675, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00033, + "step": 395, + "tokens/total": 11952672, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 180762 + }, + { + "epoch": 1.546875, + "grad_norm": 0.14256854355335236, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00286297220736742, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00287, + "step": 396, + "tokens/total": 11983088, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 181226 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.00555575592443347, + "learning_rate": 2.202206960760984e-05, + "loss": 9.771026816451922e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0001, + "step": 397, + "tokens/total": 12013488, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 181714 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.007696876302361488, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0001580697571625933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00016, + "step": 398, + "tokens/total": 12043968, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 182142 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.0037960875779390335, + "learning_rate": 2.1629798596457056e-05, + "loss": 7.467059913324192e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 399, + "tokens/total": 12074160, + "tokens/train_per_sec_per_gpu": 34.82, + "tokens/trainable": 182604 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004514187108725309, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.00011837383499369025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00012, + "step": 400, + "tokens/total": 12104352, + "tokens/train_per_sec_per_gpu": 25.26, + "tokens/trainable": 183011 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.09468130767345428, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0007745609618723392, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00077, + "step": 401, + "tokens/total": 12134240, + "tokens/train_per_sec_per_gpu": 31.75, + "tokens/trainable": 183459 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.017941495403647423, + "learning_rate": 2.105182715082638e-05, + "loss": 0.00024286848201882094, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 402, + "tokens/total": 12164480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 183959 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.005621996242552996, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.00012180282647022977, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00012, + "step": 403, + "tokens/total": 12194640, + "tokens/train_per_sec_per_gpu": 32.46, + "tokens/trainable": 184431 + }, + { + "epoch": 1.578125, + "grad_norm": 0.004712392576038837, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010195293725701049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 404, + "tokens/total": 12224960, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 184867 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.023734599351882935, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.00040538967004977167, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00041, + "step": 405, + "tokens/total": 12255008, + "tokens/train_per_sec_per_gpu": 29.33, + "tokens/trainable": 185300 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.06495791673660278, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.000537894491571933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00054, + "step": 406, + "tokens/total": 12285248, + "tokens/train_per_sec_per_gpu": 31.47, + "tokens/trainable": 185754 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.003490478964522481, + "learning_rate": 2.011689980574966e-05, + "loss": 6.170629058033228e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00006, + "step": 407, + "tokens/total": 12315552, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 186213 + }, + { + "epoch": 1.59375, + "grad_norm": 0.005349478218704462, + "learning_rate": 1.993423839463052e-05, + "loss": 9.109645179705694e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00009, + "step": 408, + "tokens/total": 12345904, + "tokens/train_per_sec_per_gpu": 32.78, + "tokens/trainable": 186659 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011329016648232937, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00016253325156867504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12376336, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 187083 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.019948504865169525, + "learning_rate": 1.957330080137385e-05, + "loss": 0.00013768397911917418, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 410, + "tokens/total": 12406880, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 187530 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.003809648798778653, + "learning_rate": 1.9395039639322864e-05, + "loss": 5.0881026254501194e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 411, + "tokens/total": 12437088, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 187972 + }, + { + "epoch": 1.609375, + "grad_norm": 0.012643440626561642, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0001802633050829172, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00018, + "step": 412, + "tokens/total": 12465440, + "tokens/train_per_sec_per_gpu": 39.92, + "tokens/trainable": 188417 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.22178205847740173, + "learning_rate": 1.904296967493982e-05, + "loss": 0.011118180118501186, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01118, + "step": 413, + "tokens/total": 12495968, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 188868 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.0006713962648063898, + "learning_rate": 1.8869175523676064e-05, + "loss": 1.6992760720313527e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 414, + "tokens/total": 12526128, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 189337 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.006867737043648958, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00011090705083915964, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00011, + "step": 415, + "tokens/total": 12556256, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 189750 + }, + { + "epoch": 1.625, + "grad_norm": 0.016067150980234146, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0001882174692582339, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 416, + "tokens/total": 12586736, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 190216 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0036702004726976156, + "learning_rate": 1.8356842992397304e-05, + "loss": 4.983436519978568e-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": 12617168, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 190708 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.3933418095111847, + "learning_rate": 1.8189105812005714e-05, + "loss": 0.0050767892971634865, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00509, + "step": 418, + "tokens/total": 12647680, + "tokens/train_per_sec_per_gpu": 30.59, + "tokens/trainable": 191126 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.04893051087856293, + "learning_rate": 1.802290048317732e-05, + "loss": 0.0004771970270667225, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00048, + "step": 419, + "tokens/total": 12677952, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 191540 + }, + { + "epoch": 1.640625, + "grad_norm": 0.03733966872096062, + "learning_rate": 1.785823392239424e-05, + "loss": 0.0004862284695263952, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00049, + "step": 420, + "tokens/total": 12708416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.00783330574631691, + "learning_rate": 1.7695112982104225e-05, + "loss": 7.455339073203504e-05, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 421, + "tokens/total": 12738640, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 192463 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.052442099899053574, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0009662319207563996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00097, + "step": 422, + "tokens/total": 12769232, + "tokens/train_per_sec_per_gpu": 32.64, + "tokens/trainable": 192918 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.024694403633475304, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.00027854571817442775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00028, + "step": 423, + "tokens/total": 12799648, + "tokens/train_per_sec_per_gpu": 37.36, + "tokens/trainable": 193425 + }, + { + "epoch": 1.65625, + "grad_norm": 0.002282459754496813, + "learning_rate": 1.721509144218405e-05, + "loss": 5.6570435845060274e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 424, + "tokens/total": 12829920, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 193898 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.28251373767852783, + "learning_rate": 1.705822021773101e-05, + "loss": 0.010694734752178192, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01075, + "step": 425, + "tokens/total": 12860112, + "tokens/train_per_sec_per_gpu": 29.22, + "tokens/trainable": 194315 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.14038732647895813, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0015529554802924395, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00155, + "step": 426, + "tokens/total": 12890320, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 194796 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.0007188008166849613, + "learning_rate": 1.6749220968158415e-05, + "loss": 1.8845212252927013e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00002, + "step": 427, + "tokens/total": 12920656, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 195262 + }, + { + "epoch": 1.671875, + "grad_norm": 0.002983702579513192, + "learning_rate": 1.659710580175893e-05, + "loss": 5.061700358055532e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 428, + "tokens/total": 12951040, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 195680 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.00501316599547863, + "learning_rate": 1.644658873654133e-05, + "loss": 8.430491288891062e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00008, + "step": 429, + "tokens/total": 12981232, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 196145 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.6195209622383118, + "learning_rate": 1.629767603613508e-05, + "loss": 0.0007395793218165636, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00074, + "step": 430, + "tokens/total": 13011616, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 196626 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.32764554023742676, + "learning_rate": 1.615037389740547e-05, + "loss": 0.012379031628370285, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.01246, + "step": 431, + "tokens/total": 13042208, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 197067 + }, + { + "epoch": 1.6875, + "grad_norm": 0.28968432545661926, + "learning_rate": 1.600468845019576e-05, + "loss": 0.003729558549821377, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00374, + "step": 432, + "tokens/total": 13072800, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 197565 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.001470765215344727, + "learning_rate": 1.5860625757072092e-05, + "loss": 3.539394310791977e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00004, + "step": 433, + "tokens/total": 13103328, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 198015 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.019142981618642807, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00024474930251017213, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 434, + "tokens/total": 13133472, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 198442 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.009595178067684174, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0001564246922498569, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00016, + "step": 435, + "tokens/total": 13164064, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 198914 + }, + { + "epoch": 1.703125, + "grad_norm": 0.22099021077156067, + "learning_rate": 1.543823381344311e-05, + "loss": 0.014420191757380962, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01452, + "step": 436, + "tokens/total": 13194464, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 199355 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.1727798581123352, + "learning_rate": 1.5300721408011114e-05, + "loss": 0.008166543208062649, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0082, + "step": 437, + "tokens/total": 13224992, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 199830 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.015094809234142303, + "learning_rate": 1.5164861051607254e-05, + "loss": 0.0002187314530601725, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 438, + "tokens/total": 13255296, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 200338 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.07529865950345993, + "learning_rate": 1.5030658397935521e-05, + "loss": 0.0006286898860707879, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00063, + "step": 439, + "tokens/total": 13285344, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 200785 + }, + { + "epoch": 1.71875, + "grad_norm": 0.005869010463356972, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00010347182978875935, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 440, + "tokens/total": 13315632, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 201267 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.02751794271171093, + "learning_rate": 1.476724846845306e-05, + "loss": 0.000567559851333499, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00057, + "step": 441, + "tokens/total": 13346048, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 201734 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.06305643171072006, + "learning_rate": 1.463805215420471e-05, + "loss": 0.0012271346058696508, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00123, + "step": 442, + "tokens/total": 13376304, + "tokens/train_per_sec_per_gpu": 30.3, + "tokens/trainable": 202174 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.008338473737239838, + "learning_rate": 1.451053546535705e-05, + "loss": 0.00016968630370683968, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00017, + "step": 443, + "tokens/total": 13406720, + "tokens/train_per_sec_per_gpu": 36.21, + "tokens/trainable": 202612 + }, + { + "epoch": 1.734375, + "grad_norm": 0.020954634994268417, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0005087319877929986, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00051, + "step": 444, + "tokens/total": 13436464, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 203020 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.019224021583795547, + "learning_rate": 1.4260562119706606e-05, + "loss": 0.0003236978664062917, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00032, + "step": 445, + "tokens/total": 13466672, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 203479 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.02892594039440155, + "learning_rate": 1.413811586531508e-05, + "loss": 0.0004480727366171777, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00045, + "step": 446, + "tokens/total": 13496736, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 203946 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.013984446413815022, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.00018316772184334695, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00018, + "step": 447, + "tokens/total": 13526928, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 204382 + }, + { + "epoch": 1.75, + "grad_norm": 0.009344048798084259, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.00011864845873787999, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00012, + "step": 448, + "tokens/total": 13557392, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 204830 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.022271448746323586, + "learning_rate": 1.3780999708818058e-05, + "loss": 0.00033031520433723927, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00033, + "step": 449, + "tokens/total": 13587776, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 205300 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.3953968286514282, + "learning_rate": 1.3665385037857758e-05, + "loss": 0.0029637760017067194, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00297, + "step": 450, + "tokens/total": 13618112, + "tokens/train_per_sec_per_gpu": 30.65, + "tokens/trainable": 205726 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.008145474828779697, + "learning_rate": 1.3551490468947126e-05, + "loss": 0.00017418013885617256, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00017, + "step": 451, + "tokens/total": 13648512, + "tokens/train_per_sec_per_gpu": 31.66, + "tokens/trainable": 206163 + }, + { + "epoch": 1.765625, + "grad_norm": 0.18700900673866272, + "learning_rate": 1.3439320741704075e-05, + "loss": 0.011147091165184975, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01121, + "step": 452, + "tokens/total": 13678704, + "tokens/train_per_sec_per_gpu": 35.39, + "tokens/trainable": 206619 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.2209583818912506, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.004581573884934187, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00459, + "step": 453, + "tokens/total": 13709232, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 207101 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.04667079448699951, + "learning_rate": 1.3220174411609587e-05, + "loss": 0.0004827629018109292, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00048, + "step": 454, + "tokens/total": 13739600, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 207546 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.22248415648937225, + "learning_rate": 1.3113206928331471e-05, + "loss": 0.003234278876334429, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00324, + "step": 455, + "tokens/total": 13769936, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 207989 + }, + { + "epoch": 1.78125, + "grad_norm": 0.006573830731213093, + "learning_rate": 1.300798252548806e-05, + "loss": 0.00013498595217242837, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00013, + "step": 456, + "tokens/total": 13800240, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 208402 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.0018037718255072832, + "learning_rate": 1.2904505581896265e-05, + "loss": 3.7187150155659765e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 457, + "tokens/total": 13830512, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 208904 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.01422063261270523, + "learning_rate": 1.2802780403654082e-05, + "loss": 0.00024471062351949513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00024, + "step": 458, + "tokens/total": 13860832, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 209382 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.01398254744708538, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.000272503646556288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00027, + "step": 459, + "tokens/total": 13890992, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 209833 + }, + { + "epoch": 1.796875, + "grad_norm": 0.12035191804170609, + "learning_rate": 1.2604602202943861e-05, + "loss": 0.0023982198908925056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0024, + "step": 460, + "tokens/total": 13921424, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 210284 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.025684842839837074, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.0005578990676440299, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00056, + "step": 461, + "tokens/total": 13951504, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 210769 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.0697178840637207, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.0008336097234860063, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00083, + "step": 462, + "tokens/total": 13979728, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 211193 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.0039046104066073895, + "learning_rate": 1.2320576593470082e-05, + "loss": 9.185953240375966e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00009, + "step": 463, + "tokens/total": 14009936, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 211655 + }, + { + "epoch": 1.8125, + "grad_norm": 0.007744460366666317, + "learning_rate": 1.2229448340928828e-05, + "loss": 0.00017765363736543804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00018, + "step": 464, + "tokens/total": 14039872, + "tokens/train_per_sec_per_gpu": 29.3, + "tokens/trainable": 212085 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.03343091905117035, + "learning_rate": 1.2140099945624458e-05, + "loss": 0.00046598759945482016, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00047, + "step": 465, + "tokens/total": 14070096, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 212570 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.48290523886680603, + "learning_rate": 1.205253512570841e-05, + "loss": 0.016730966046452522, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01687, + "step": 466, + "tokens/total": 14100192, + "tokens/train_per_sec_per_gpu": 32.9, + "tokens/trainable": 213011 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.05692508816719055, + "learning_rate": 1.1966757525110255e-05, + "loss": 0.0009399584960192442, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00094, + "step": 467, + "tokens/total": 14130448, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 213450 + }, + { + "epoch": 1.828125, + "grad_norm": 0.008330056443810463, + "learning_rate": 1.1882770713386095e-05, + "loss": 0.0001491767616244033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00015, + "step": 468, + "tokens/total": 14160480, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 213898 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.005805026274174452, + "learning_rate": 1.180057818556998e-05, + "loss": 9.818439139053226e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0001, + "step": 469, + "tokens/total": 14191152, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 214366 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.005398476962000132, + "learning_rate": 1.1720183362028494e-05, + "loss": 0.00013144969125278294, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00013, + "step": 470, + "tokens/total": 14219392, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 214800 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.006290988065302372, + "learning_rate": 1.1641589588318387e-05, + "loss": 0.00014237320283427835, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00014, + "step": 471, + "tokens/total": 14249920, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 215292 + }, + { + "epoch": 1.84375, + "grad_norm": 0.0057176705449819565, + "learning_rate": 1.1564800135047418e-05, + "loss": 0.00012321076064836234, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00012, + "step": 472, + "tokens/total": 14280048, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 215741 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.03017423115670681, + "learning_rate": 1.148981819773816e-05, + "loss": 0.0006377531681209803, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00064, + "step": 473, + "tokens/total": 14310432, + "tokens/train_per_sec_per_gpu": 34.63, + "tokens/trainable": 216194 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.04882626235485077, + "learning_rate": 1.1416646896695086e-05, + "loss": 0.0009199330816045403, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00092, + "step": 474, + "tokens/total": 14340496, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 216634 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.49993258714675903, + "learning_rate": 1.1345289276874717e-05, + "loss": 0.0011110955383628607, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00111, + "step": 475, + "tokens/total": 14370720, + "tokens/train_per_sec_per_gpu": 30.46, + "tokens/trainable": 217040 + }, + { + "epoch": 1.859375, + "grad_norm": 0.015277047641575336, + "learning_rate": 1.1275748307758873e-05, + "loss": 0.0002883652050513774, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00029, + "step": 476, + "tokens/total": 14401136, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 217530 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.05111508071422577, + "learning_rate": 1.1208026883231147e-05, + "loss": 0.0007504059467464685, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00075, + "step": 477, + "tokens/total": 14431360, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 217988 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.11643723398447037, + "learning_rate": 1.1142127821456433e-05, + "loss": 0.0013695022789761424, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00137, + "step": 478, + "tokens/total": 14461968, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 218461 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.027246320620179176, + "learning_rate": 1.1078053864763674e-05, + "loss": 0.0004469662962947041, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00045, + "step": 479, + "tokens/total": 14491984, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 218909 + }, + { + "epoch": 1.875, + "grad_norm": 0.02049543894827366, + "learning_rate": 1.1015807679531756e-05, + "loss": 0.00044887623516842723, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00045, + "step": 480, + "tokens/total": 14522544, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 219350 + } + ], + "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": 9.857299273093647e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-480/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-512/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-512/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-512/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b9d7227c4027e9061b987572a62e5abf8e2c252d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:020c719595b27274ab926638abacda614a046afdcd79f4e60b555a6c403dae8d +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-512/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..73360eb5201ab03281a8df0bd923fd16a58ff76c --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0012fb39a15a5797be5f5f8d34f79bdf3127d7037add728dde1bf8acc28355c +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..93906269157a5485207c9bb4b3b757d5ea700afa --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75e98dcceb68f1b6ee34a96307625206231cf5a0e0e4e16d8a9eafccc33ea3fe +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..22325b3b3d08e0697784fe21b8c321485a51f7b5 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-512/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-512/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-512/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..242dd434cbe768e1e6c671a04c0c09f21072a0fc --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/tokens_state.json @@ -0,0 +1 @@ +{"total": 15491760, "trainable": 234078} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..e3273853b4dc043946fe4f7a0dd34b341ef515a7 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.37105968594551086, + "learning_rate": 9.53619478457953e-05, + "loss": 0.007098545785993338, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00712, + "step": 97, + "tokens/total": 2936752, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 44255 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.0755594968795776, + "learning_rate": 9.523275153154695e-05, + "loss": 0.028615131974220276, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02903, + "step": 98, + "tokens/total": 2967072, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 44739 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.9296021461486816, + "learning_rate": 9.51018809682839e-05, + "loss": 0.01639867201447487, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01653, + "step": 99, + "tokens/total": 2997664, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 45202 + }, + { + "epoch": 0.390625, + "grad_norm": 0.6913011074066162, + "learning_rate": 9.49693416020645e-05, + "loss": 0.015926485881209373, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01605, + "step": 100, + "tokens/total": 3028032, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 45646 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3258785605430603, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008693980053067207, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00873, + "step": 101, + "tokens/total": 3058272, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 46089 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.0787935256958008, + "learning_rate": 9.469927859198888e-05, + "loss": 0.01954980194568634, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01974, + "step": 102, + "tokens/total": 3088848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 46547 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.9061365127563477, + "learning_rate": 9.456176618655689e-05, + "loss": 0.03529620170593262, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03593, + "step": 103, + "tokens/total": 3119424, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 47014 + }, + { + "epoch": 0.40625, + "grad_norm": 2.671016216278076, + "learning_rate": 9.442260745454927e-05, + "loss": 0.02147374488413334, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02171, + "step": 104, + "tokens/total": 3149696, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 47503 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.26923975348472595, + "learning_rate": 9.428180818692884e-05, + "loss": 0.007999668829143047, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00803, + "step": 105, + "tokens/total": 3180064, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 47996 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.7145718932151794, + "learning_rate": 9.413937424292791e-05, + "loss": 0.030988795682787895, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.03147, + "step": 106, + "tokens/total": 3208320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 48447 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.20594321191310883, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007135331630706787, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00716, + "step": 107, + "tokens/total": 3238416, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 48889 + }, + { + "epoch": 0.421875, + "grad_norm": 0.2812410891056061, + "learning_rate": 9.384962610259455e-05, + "loss": 0.010571298189461231, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01063, + "step": 108, + "tokens/total": 3268784, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 49360 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.3485014736652374, + "learning_rate": 9.370232396386494e-05, + "loss": 0.016001982614398003, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01613, + "step": 109, + "tokens/total": 3298912, + "tokens/train_per_sec_per_gpu": 36.73, + "tokens/trainable": 49827 + }, + { + "epoch": 0.4296875, + "grad_norm": 1.474901795387268, + "learning_rate": 9.355341126345868e-05, + "loss": 0.024302365258336067, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.0246, + "step": 110, + "tokens/total": 3329232, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 50267 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.5579241514205933, + "learning_rate": 9.340289419824107e-05, + "loss": 0.017602810636162758, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01776, + "step": 111, + "tokens/total": 3359440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 50711 + }, + { + "epoch": 0.4375, + "grad_norm": 0.5616764426231384, + "learning_rate": 9.325077903184159e-05, + "loss": 0.015218488872051239, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01533, + "step": 112, + "tokens/total": 3389472, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 51157 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.38610321283340454, + "learning_rate": 9.30970720943932e-05, + "loss": 0.008289901539683342, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00832, + "step": 113, + "tokens/total": 3419920, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 51613 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.3681856095790863, + "learning_rate": 9.2941779782269e-05, + "loss": 0.022977739572525024, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02324, + "step": 114, + "tokens/total": 3450176, + "tokens/train_per_sec_per_gpu": 29.07, + "tokens/trainable": 52017 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.2453298270702362, + "learning_rate": 9.278490855781596e-05, + "loss": 0.004594211932271719, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0046, + "step": 115, + "tokens/total": 3480544, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 52506 + }, + { + "epoch": 0.453125, + "grad_norm": 0.37761810421943665, + "learning_rate": 9.262646494908604e-05, + "loss": 0.011891474947333336, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01196, + "step": 116, + "tokens/total": 3510848, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 53008 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.953158974647522, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006801537703722715, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00682, + "step": 117, + "tokens/total": 3541296, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 53467 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.45078471302986145, + "learning_rate": 9.230488701789578e-05, + "loss": 0.017124727368354797, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01727, + "step": 118, + "tokens/total": 3571936, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 53923 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.6464744806289673, + "learning_rate": 9.214176607760577e-05, + "loss": 0.012542763724923134, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01262, + "step": 119, + "tokens/total": 3602000, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 54363 + }, + { + "epoch": 0.46875, + "grad_norm": 0.3361028730869293, + "learning_rate": 9.197709951682268e-05, + "loss": 0.004235772415995598, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00424, + "step": 120, + "tokens/total": 3632288, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 54839 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4455215036869049, + "learning_rate": 9.181089418799428e-05, + "loss": 0.004468773491680622, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 121, + "tokens/total": 3662608, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 55267 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3547821044921875, + "learning_rate": 9.164315700760271e-05, + "loss": 0.010471776127815247, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01053, + "step": 122, + "tokens/total": 3692864, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 55700 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7816572189331055, + "learning_rate": 9.147389495587671e-05, + "loss": 0.01556549035012722, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01569, + "step": 123, + "tokens/total": 3722864, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 56144 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3188917338848114, + "learning_rate": 9.130311507650116e-05, + "loss": 0.0064091756939888, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00643, + "step": 124, + "tokens/total": 3753056, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 56589 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.7098766565322876, + "learning_rate": 9.113082447632394e-05, + "loss": 0.026026558130979538, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02637, + "step": 125, + "tokens/total": 3783312, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 57081 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4195394217967987, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01531171053647995, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01543, + "step": 126, + "tokens/total": 3813712, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 57527 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.4466935992240906, + "learning_rate": 9.078173985499394e-05, + "loss": 0.014355337247252464, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01446, + "step": 127, + "tokens/total": 3844336, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 58005 + }, + { + "epoch": 0.5, + "grad_norm": 0.19709816575050354, + "learning_rate": 9.060496036067713e-05, + "loss": 0.006424291059374809, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00644, + "step": 128, + "tokens/total": 3872688, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 58454 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.18546900153160095, + "learning_rate": 9.042669919862615e-05, + "loss": 0.003899412928149104, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00391, + "step": 129, + "tokens/total": 3902640, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 58896 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2095804512500763, + "learning_rate": 9.024696378701557e-05, + "loss": 0.005734554957598448, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00575, + "step": 130, + "tokens/total": 3933104, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 59356 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.3548737168312073, + "learning_rate": 9.006576160536948e-05, + "loss": 0.010866682976484299, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01093, + "step": 131, + "tokens/total": 3963184, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 59792 + }, + { + "epoch": 0.515625, + "grad_norm": 0.24394433200359344, + "learning_rate": 8.988310019425035e-05, + "loss": 0.004735369700938463, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00475, + "step": 132, + "tokens/total": 3993648, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 60266 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.3678644597530365, + "learning_rate": 8.969898715494506e-05, + "loss": 0.007982447743415833, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00801, + "step": 133, + "tokens/total": 4024000, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 60759 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.6605305075645447, + "learning_rate": 8.951343014914869e-05, + "loss": 0.02931230701506138, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.02975, + "step": 134, + "tokens/total": 4052480, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 61204 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.1193099245429039, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0025630262680351734, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00257, + "step": 135, + "tokens/total": 4082864, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 61684 + }, + { + "epoch": 0.53125, + "grad_norm": 0.12960873544216156, + "learning_rate": 8.913801518498845e-05, + "loss": 0.0019581567030400038, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00196, + "step": 136, + "tokens/total": 4113040, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 62159 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.34460127353668213, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01051283534616232, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01057, + "step": 137, + "tokens/total": 4143344, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 62649 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.20421279966831207, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0035965177230536938, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0036, + "step": 138, + "tokens/total": 4173648, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 63100 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.30597925186157227, + "learning_rate": 8.856425797031829e-05, + "loss": 0.008190772496163845, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00822, + "step": 139, + "tokens/total": 4204080, + "tokens/train_per_sec_per_gpu": 37.61, + "tokens/trainable": 63588 + }, + { + "epoch": 0.546875, + "grad_norm": 0.5688289999961853, + "learning_rate": 8.837020140354295e-05, + "loss": 0.022904174402356148, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.02317, + "step": 140, + "tokens/total": 4232400, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 64005 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.5879082083702087, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013611489906907082, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0137, + "step": 141, + "tokens/total": 4262624, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 64462 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.31034204363822937, + "learning_rate": 8.797793039239017e-05, + "loss": 0.008315861225128174, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00835, + "step": 142, + "tokens/total": 4293168, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 64900 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.7010869979858398, + "learning_rate": 8.777973227201069e-05, + "loss": 0.022431056946516037, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02268, + "step": 143, + "tokens/total": 4323248, + "tokens/train_per_sec_per_gpu": 37.57, + "tokens/trainable": 65379 + }, + { + "epoch": 0.5625, + "grad_norm": 0.2407914400100708, + "learning_rate": 8.758017005316988e-05, + "loss": 0.006207741796970367, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 144, + "tokens/total": 4353344, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 65845 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21913351118564606, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009031139314174652, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00907, + "step": 145, + "tokens/total": 4383584, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 66285 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.39635229110717773, + "learning_rate": 8.717698659491851e-05, + "loss": 0.016251137480139732, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01638, + "step": 146, + "tokens/total": 4413856, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 66736 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.5958674550056458, + "learning_rate": 8.697338213361735e-05, + "loss": 0.018637506291270256, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01881, + "step": 147, + "tokens/total": 4444112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 67150 + }, + { + "epoch": 0.578125, + "grad_norm": 0.19332216680049896, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006893862038850784, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00692, + "step": 148, + "tokens/total": 4474464, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 67633 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.36499589681625366, + "learning_rate": 8.656219011037509e-05, + "loss": 0.010607698000967503, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01066, + "step": 149, + "tokens/total": 4504880, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 68077 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.10280518978834152, + "learning_rate": 8.63546196598125e-05, + "loss": 0.002519096015021205, + "memory/device_reserved (GiB)": 35.79, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00252, + "step": 150, + "tokens/total": 4534976, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 68545 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.24021312594413757, + "learning_rate": 8.614574441554145e-05, + "loss": 0.006152496673166752, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00617, + "step": 151, + "tokens/total": 4565088, + "tokens/train_per_sec_per_gpu": 30.26, + "tokens/trainable": 69011 + }, + { + "epoch": 0.59375, + "grad_norm": 0.5905681252479553, + "learning_rate": 8.593557306971349e-05, + "loss": 0.0134108392521739, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0135, + "step": 152, + "tokens/total": 4595600, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 69465 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5932232141494751, + "learning_rate": 8.572411436841618e-05, + "loss": 0.03526075929403305, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03589, + "step": 153, + "tokens/total": 4626032, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 69928 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.21574290096759796, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0023245825432240963, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00233, + "step": 154, + "tokens/total": 4656352, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 70395 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.08898661285638809, + "learning_rate": 8.529737015125824e-05, + "loss": 0.0007986892596818507, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0008, + "step": 155, + "tokens/total": 4684672, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 70829 + }, + { + "epoch": 0.609375, + "grad_norm": 0.1178978830575943, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0014696650905534625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00147, + "step": 156, + "tokens/total": 4715216, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 71287 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06257754564285278, + "learning_rate": 8.486558279760375e-05, + "loss": 0.000776763423345983, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00078, + "step": 157, + "tokens/total": 4745376, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 71749 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.8163263201713562, + "learning_rate": 8.464782037243449e-05, + "loss": 0.033990029245615005, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03457, + "step": 158, + "tokens/total": 4775776, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 72185 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.014705101028084755, + "learning_rate": 8.442882418044202e-05, + "loss": 0.00028892228147014976, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00029, + "step": 159, + "tokens/total": 4805856, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 72604 + }, + { + "epoch": 0.625, + "grad_norm": 0.4126265048980713, + "learning_rate": 8.420860333495179e-05, + "loss": 0.019177839159965515, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01936, + "step": 160, + "tokens/total": 4836304, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 73061 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.6193662285804749, + "learning_rate": 8.398716700025208e-05, + "loss": 0.008987972512841225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00903, + "step": 161, + "tokens/total": 4866544, + "tokens/train_per_sec_per_gpu": 29.97, + "tokens/trainable": 73474 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.6351296901702881, + "learning_rate": 8.376452439121266e-05, + "loss": 0.01764395274221897, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0178, + "step": 162, + "tokens/total": 4897120, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 73923 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.19809548556804657, + "learning_rate": 8.354068477290124e-05, + "loss": 0.002400235738605261, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0024, + "step": 163, + "tokens/total": 4927440, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 74379 + }, + { + "epoch": 0.640625, + "grad_norm": 0.17828311026096344, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0019721905700862408, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00197, + "step": 164, + "tokens/total": 4958032, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 74830 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.14491626620292664, + "learning_rate": 8.308945181740812e-05, + "loss": 0.0023866910487413406, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00239, + "step": 165, + "tokens/total": 4988368, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 75262 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.11567407846450806, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0012497061397880316, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00125, + "step": 166, + "tokens/total": 5018672, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 75703 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.3614877164363861, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010097509250044823, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01015, + "step": 167, + "tokens/total": 5049344, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 76164 + }, + { + "epoch": 0.65625, + "grad_norm": 0.21049293875694275, + "learning_rate": 8.240385928474219e-05, + "loss": 0.005174943245947361, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00519, + "step": 168, + "tokens/total": 5079856, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 76623 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.16444046795368195, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0028489381074905396, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00285, + "step": 169, + "tokens/total": 5110192, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 77047 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.29660263657569885, + "learning_rate": 8.194107981329746e-05, + "loss": 0.020341932773590088, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02055, + "step": 170, + "tokens/total": 5140592, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 77496 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.2912541627883911, + "learning_rate": 8.170800355882518e-05, + "loss": 0.013131296262145042, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01322, + "step": 171, + "tokens/total": 5170848, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 77956 + }, + { + "epoch": 0.671875, + "grad_norm": 0.14199896156787872, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002253658836707473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00226, + "step": 172, + "tokens/total": 5201280, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 78410 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.6508966684341431, + "learning_rate": 8.123852650824877e-05, + "loss": 0.01731065846979618, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01746, + "step": 173, + "tokens/total": 5231712, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 78849 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.22890180349349976, + "learning_rate": 8.100214524900103e-05, + "loss": 0.013438408263027668, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.01353, + "step": 174, + "tokens/total": 5261440, + "tokens/train_per_sec_per_gpu": 30.1, + "tokens/trainable": 79266 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.034085191786289215, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0010016995947808027, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.001, + "step": 175, + "tokens/total": 5291728, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 79736 + }, + { + "epoch": 0.6875, + "grad_norm": 0.1346512883901596, + "learning_rate": 8.052614644612253e-05, + "loss": 0.0026639762800186872, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00267, + "step": 176, + "tokens/total": 5322240, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 80214 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.2848314940929413, + "learning_rate": 8.028654871074489e-05, + "loss": 0.008686527609825134, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00872, + "step": 177, + "tokens/total": 5352576, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 80661 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.037947915494441986, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0010221066186204553, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00102, + "step": 178, + "tokens/total": 5382784, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 81129 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.1301245242357254, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017726586665958166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00177, + "step": 179, + "tokens/total": 5412784, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 81576 + }, + { + "epoch": 0.703125, + "grad_norm": 0.16076700389385223, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0021704614628106356, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 180, + "tokens/total": 5442960, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 82049 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.5916904807090759, + "learning_rate": 7.931773536489872e-05, + "loss": 0.018786190077662468, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01896, + "step": 181, + "tokens/total": 5473216, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 82566 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.140871062874794, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002907736459746957, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00291, + "step": 182, + "tokens/total": 5503568, + "tokens/train_per_sec_per_gpu": 33.93, + "tokens/trainable": 83054 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3186001181602478, + "learning_rate": 7.882721650609442e-05, + "loss": 0.002825568662956357, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00283, + "step": 183, + "tokens/total": 5533760, + "tokens/train_per_sec_per_gpu": 32.79, + "tokens/trainable": 83501 + }, + { + "epoch": 0.71875, + "grad_norm": 0.10091359168291092, + "learning_rate": 7.85804646415409e-05, + "loss": 0.000788951525464654, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00079, + "step": 184, + "tokens/total": 5564240, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 84007 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.29195457696914673, + "learning_rate": 7.833273149760207e-05, + "loss": 0.006313872057944536, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00633, + "step": 185, + "tokens/total": 5594448, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 84491 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.5876129269599915, + "learning_rate": 7.808402738346527e-05, + "loss": 0.01837407797574997, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01854, + "step": 186, + "tokens/total": 5624816, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 84952 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.4564031660556793, + "learning_rate": 7.783436264872382e-05, + "loss": 0.00939339492470026, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00944, + "step": 187, + "tokens/total": 5655104, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 85400 + }, + { + "epoch": 0.734375, + "grad_norm": 0.018894854933023453, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0005394529434852302, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00054, + "step": 188, + "tokens/total": 5685360, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 85889 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.11897671222686768, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0019742052536457777, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00198, + "step": 189, + "tokens/total": 5715648, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 86396 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.24015508592128754, + "learning_rate": 7.707970881383977e-05, + "loss": 0.007755734957754612, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00779, + "step": 190, + "tokens/total": 5746240, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 86867 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.2647373378276825, + "learning_rate": 7.682630588562518e-05, + "loss": 0.017713403329253197, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.01787, + "step": 191, + "tokens/total": 5776816, + "tokens/train_per_sec_per_gpu": 36.16, + "tokens/trainable": 87314 + }, + { + "epoch": 0.75, + "grad_norm": 0.2803303599357605, + "learning_rate": 7.657199467573129e-05, + "loss": 0.008597557432949543, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00863, + "step": 192, + "tokens/total": 5807248, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 87736 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.045643068850040436, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008268685778602958, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 193, + "tokens/total": 5837680, + "tokens/train_per_sec_per_gpu": 35.05, + "tokens/trainable": 88219 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.49878475069999695, + "learning_rate": 7.606068977997255e-05, + "loss": 0.008574062027037144, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00861, + "step": 194, + "tokens/total": 5868000, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 88699 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.0975877121090889, + "learning_rate": 7.580371737159148e-05, + "loss": 0.0020896620117127895, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00209, + "step": 195, + "tokens/total": 5896224, + "tokens/train_per_sec_per_gpu": 30.38, + "tokens/trainable": 89132 + }, + { + "epoch": 0.765625, + "grad_norm": 0.18829824030399323, + "learning_rate": 7.554587923561324e-05, + "loss": 0.004770598374307156, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00478, + "step": 196, + "tokens/total": 5926624, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 89594 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.11453713476657867, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0024538300931453705, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00246, + "step": 197, + "tokens/total": 5954944, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 90042 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.3371991217136383, + "learning_rate": 7.502764873523431e-05, + "loss": 0.012420275248587132, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0125, + "step": 198, + "tokens/total": 5985200, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 90503 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.044131435453891754, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0006442728918045759, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00064, + "step": 199, + "tokens/total": 6015952, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 90982 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11669042706489563, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0016041958006098866, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 200, + "tokens/total": 6046208, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 91466 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.030662450939416885, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0007836977019906044, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00078, + "step": 201, + "tokens/total": 6076208, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 91912 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.16404050588607788, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0035354122519493103, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00354, + "step": 202, + "tokens/total": 6106560, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 92385 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.3355195224285126, + "learning_rate": 7.371767763212238e-05, + "loss": 0.029696868732571602, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.03014, + "step": 203, + "tokens/total": 6136640, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 92842 + }, + { + "epoch": 0.796875, + "grad_norm": 0.2913725972175598, + "learning_rate": 7.345330287655617e-05, + "loss": 0.004002463538199663, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00401, + "step": 204, + "tokens/total": 6167184, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 93323 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.43749040365219116, + "learning_rate": 7.31881602037339e-05, + "loss": 0.013340720906853676, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01343, + "step": 205, + "tokens/total": 6197376, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 93756 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.4968360960483551, + "learning_rate": 7.29222606473245e-05, + "loss": 0.01012748945504427, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01018, + "step": 206, + "tokens/total": 6227600, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 94235 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.06706669926643372, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0005324441008269787, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00053, + "step": 207, + "tokens/total": 6257760, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 94733 + }, + { + "epoch": 0.8125, + "grad_norm": 0.14510789513587952, + "learning_rate": 7.238823517544436e-05, + "loss": 0.0053801448084414005, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00539, + "step": 208, + "tokens/total": 6288048, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 95218 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.07251987606287003, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0016005284851416945, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0016, + "step": 209, + "tokens/total": 6318688, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 95619 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.026614652946591377, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0005790021969005466, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00058, + "step": 210, + "tokens/total": 6348944, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 96076 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.04637908935546875, + "learning_rate": 7.158179796885005e-05, + "loss": 0.0009742251713760197, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00097, + "step": 211, + "tokens/total": 6378960, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 96505 + }, + { + "epoch": 0.828125, + "grad_norm": 0.07815160602331161, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0010261747520416975, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00103, + "step": 212, + "tokens/total": 6409472, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 96930 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.032491739839315414, + "learning_rate": 7.104070433827139e-05, + "loss": 0.0007142109097912908, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00071, + "step": 213, + "tokens/total": 6439904, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 97361 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.26459500193595886, + "learning_rate": 7.076915060786705e-05, + "loss": 0.00510193407535553, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00511, + "step": 214, + "tokens/total": 6470176, + "tokens/train_per_sec_per_gpu": 36.36, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4014389216899872, + "learning_rate": 7.049694065873882e-05, + "loss": 0.01917858049273491, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01936, + "step": 215, + "tokens/total": 6500336, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.017172295600175858, + "learning_rate": 7.022408581865382e-05, + "loss": 0.00041207089088857174, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00041, + "step": 216, + "tokens/total": 6530688, + "tokens/train_per_sec_per_gpu": 36.38, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.21681398153305054, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0024940790608525276, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0025, + "step": 217, + "tokens/total": 6561040, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.15546201169490814, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0021202065981924534, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00212, + "step": 218, + "tokens/total": 6589392, + "tokens/train_per_sec_per_gpu": 39.67, + "tokens/trainable": 99731 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.14469008147716522, + "learning_rate": 6.940176563004123e-05, + "loss": 0.001545824226923287, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00155, + "step": 219, + "tokens/total": 6619824, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 100164 + }, + { + "epoch": 0.859375, + "grad_norm": 0.36150020360946655, + "learning_rate": 6.912644503343682e-05, + "loss": 0.015608995221555233, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01573, + "step": 220, + "tokens/total": 6650224, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 100620 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.05116768181324005, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0009821112034842372, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00098, + "step": 221, + "tokens/total": 6680384, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 101080 + }, + { + "epoch": 0.8671875, + "grad_norm": 1.1284459829330444, + "learning_rate": 6.857405174478604e-05, + "loss": 0.016592765226960182, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01673, + "step": 222, + "tokens/total": 6710896, + "tokens/train_per_sec_per_gpu": 37.76, + "tokens/trainable": 101546 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.0270449947565794, + "learning_rate": 6.82970020400792e-05, + "loss": 0.0005479598185047507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00055, + "step": 223, + "tokens/total": 6741280, + "tokens/train_per_sec_per_gpu": 31.55, + "tokens/trainable": 101987 + }, + { + "epoch": 0.875, + "grad_norm": 0.28190311789512634, + "learning_rate": 6.801939899284132e-05, + "loss": 0.016690485179424286, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01683, + "step": 224, + "tokens/total": 6771488, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 102435 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.31064996123313904, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007865586318075657, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0079, + "step": 225, + "tokens/total": 6801744, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 102843 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.21806642413139343, + "learning_rate": 6.746257910210214e-05, + "loss": 0.006044697016477585, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00606, + "step": 226, + "tokens/total": 6832432, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 103342 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.09062594920396805, + "learning_rate": 6.718338543014937e-05, + "loss": 0.001718292129226029, + "memory/device_reserved (GiB)": 35.71, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00172, + "step": 227, + "tokens/total": 6862448, + "tokens/train_per_sec_per_gpu": 40.76, + "tokens/trainable": 103844 + }, + { + "epoch": 0.890625, + "grad_norm": 0.9462765455245972, + "learning_rate": 6.69036847577983e-05, + "loss": 0.01898857392370701, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01917, + "step": 228, + "tokens/total": 6892832, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 104307 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.0640437975525856, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0018462492153048515, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00185, + "step": 229, + "tokens/total": 6922832, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 104772 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.11650252342224121, + "learning_rate": 6.63428089904618e-05, + "loss": 0.004365470260381699, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00438, + "step": 230, + "tokens/total": 6953104, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 105240 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.07681138068437576, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002444313373416662, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00245, + "step": 231, + "tokens/total": 6983440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 105703 + }, + { + "epoch": 0.90625, + "grad_norm": 0.20511816442012787, + "learning_rate": 6.578004516044172e-05, + "loss": 0.004132373258471489, + "memory/device_reserved (GiB)": 35.77, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00414, + "step": 232, + "tokens/total": 7013840, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 106179 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.11913148313760757, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0025868406519293785, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00259, + "step": 233, + "tokens/total": 7044240, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 106683 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03754379227757454, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0014621062437072396, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00146, + "step": 234, + "tokens/total": 7074224, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 107121 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.09151529520750046, + "learning_rate": 6.493256429322259e-05, + "loss": 0.002102814381942153, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00211, + "step": 235, + "tokens/total": 7104672, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 107561 + }, + { + "epoch": 0.921875, + "grad_norm": 0.09312661737203598, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002275613136589527, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00228, + "step": 236, + "tokens/total": 7134784, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 108018 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.02780505083501339, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0009309989400207996, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00093, + "step": 237, + "tokens/total": 7164960, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 108446 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.014633426442742348, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0004528576973825693, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00045, + "step": 238, + "tokens/total": 7195344, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 108866 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.12943324446678162, + "learning_rate": 6.379685834195036e-05, + "loss": 0.001399983186274767, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 239, + "tokens/total": 7225680, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 109316 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1406504064798355, + "learning_rate": 6.351198709240186e-05, + "loss": 0.002863554749637842, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00287, + "step": 240, + "tokens/total": 7256112, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 109787 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.016691425815224648, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0003821298014372587, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00038, + "step": 241, + "tokens/total": 7286384, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 110246 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.038080595433712006, + "learning_rate": 6.294119380711849e-05, + "loss": 0.000662593636661768, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00066, + "step": 242, + "tokens/total": 7316688, + "tokens/train_per_sec_per_gpu": 29.16, + "tokens/trainable": 110667 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2705026865005493, + "learning_rate": 6.265529552442209e-05, + "loss": 0.010765114799141884, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01082, + "step": 243, + "tokens/total": 7346992, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 111154 + }, + { + "epoch": 0.953125, + "grad_norm": 0.007892047055065632, + "learning_rate": 6.236907867363127e-05, + "loss": 0.00019494083244353533, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00019, + "step": 244, + "tokens/total": 7377632, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 111611 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.03014756739139557, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0004936520708724856, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00049, + "step": 245, + "tokens/total": 7407872, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 112102 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.17086820304393768, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0017580927815288305, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00176, + "step": 246, + "tokens/total": 7438448, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 112569 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.004760230891406536, + "learning_rate": 6.150863588251694e-05, + "loss": 0.00013790707453154027, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 247, + "tokens/total": 7466768, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 113023 + }, + { + "epoch": 0.96875, + "grad_norm": 0.02571815438568592, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0004854568687733263, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00049, + "step": 248, + "tokens/total": 7497360, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 113464 + }, + { + "epoch": 0.97265625, + "grad_norm": 5.738894462585449, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.0020567248575389385, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00206, + "step": 249, + "tokens/total": 7527504, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 113908 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.1298590898513794, + "learning_rate": 6.064575550087316e-05, + "loss": 0.001127784256823361, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00113, + "step": 250, + "tokens/total": 7555808, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 114348 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.389615923166275, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020728919189423323, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00208, + "step": 251, + "tokens/total": 7586320, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 114830 + }, + { + "epoch": 0.984375, + "grad_norm": 0.011385864578187466, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0001549161970615387, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00015, + "step": 252, + "tokens/total": 7616576, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 115278 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.041026510298252106, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0002571218938101083, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00026, + "step": 253, + "tokens/total": 7647152, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 115699 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.053424153476953506, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0005855828057974577, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00059, + "step": 254, + "tokens/total": 7677584, + "tokens/train_per_sec_per_gpu": 30.78, + "tokens/trainable": 116133 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.003189819399267435, + "learning_rate": 5.920308275189541e-05, + "loss": 6.202489021234214e-05, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7708016, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 116591 + }, + { + "epoch": 1.0, + "grad_norm": 0.024073874577879906, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.00023448803403880447, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00023, + "step": 256, + "tokens/total": 7738608, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 117039 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.37222179770469666, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.004106060601770878, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00411, + "step": 257, + "tokens/total": 7768944, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 117495 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.6812037825584412, + "learning_rate": 5.833528413179249e-05, + "loss": 0.009335177019238472, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00938, + "step": 258, + "tokens/total": 7799232, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 117925 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.003316260641440749, + "learning_rate": 5.80457242789548e-05, + "loss": 7.446527888532728e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 259, + "tokens/total": 7829568, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 118373 + }, + { + "epoch": 1.015625, + "grad_norm": 0.016280459240078926, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0001550098677398637, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00016, + "step": 260, + "tokens/total": 7860000, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 118816 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.016491129994392395, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0001946197880897671, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 261, + "tokens/total": 7890464, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 119263 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0688014104962349, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0002748112892732024, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 262, + "tokens/total": 7920944, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 119772 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.3654443919658661, + "learning_rate": 5.688633799118971e-05, + "loss": 0.005442210007458925, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00546, + "step": 263, + "tokens/total": 7951296, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 120269 + }, + { + "epoch": 1.03125, + "grad_norm": 0.11719754338264465, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0006240052171051502, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 264, + "tokens/total": 7981856, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 120746 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.042248908430337906, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.000372910697478801, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00037, + "step": 265, + "tokens/total": 8012192, + "tokens/train_per_sec_per_gpu": 38.31, + "tokens/trainable": 121204 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.20307184755802155, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0017405826365575194, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00174, + "step": 266, + "tokens/total": 8042736, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.0066534546203911304, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00011396329500712454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 267, + "tokens/total": 8073024, + "tokens/train_per_sec_per_gpu": 40.1, + "tokens/trainable": 122183 + }, + { + "epoch": 1.046875, + "grad_norm": 0.056601718068122864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0008997474797070026, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 268, + "tokens/total": 8103280, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 122645 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08372552692890167, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0007820962928235531, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00078, + "step": 269, + "tokens/total": 8133440, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 123105 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.2750205397605896, + "learning_rate": 5.485485480053015e-05, + "loss": 0.002154015703126788, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00216, + "step": 270, + "tokens/total": 8163920, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 123595 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.6561075448989868, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.006013512145727873, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00603, + "step": 271, + "tokens/total": 8194032, + "tokens/train_per_sec_per_gpu": 37.69, + "tokens/trainable": 124077 + }, + { + "epoch": 1.0625, + "grad_norm": 0.1100449338555336, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0009246943518519402, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00093, + "step": 272, + "tokens/total": 8224320, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 124547 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.026166923344135284, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030680614872835577, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00031, + "step": 273, + "tokens/total": 8254368, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 124977 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.00845333095639944, + "learning_rate": 5.369387440051111e-05, + "loss": 0.00012944119225721806, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00013, + "step": 274, + "tokens/total": 8284944, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 125438 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.5315051674842834, + "learning_rate": 5.340373499110935e-05, + "loss": 0.003738347440958023, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00375, + "step": 275, + "tokens/total": 8315584, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 125889 + }, + { + "epoch": 1.078125, + "grad_norm": 0.9425964951515198, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.005198127124458551, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00521, + "step": 276, + "tokens/total": 8345952, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 126363 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0510481633245945, + "learning_rate": 5.282366752473479e-05, + "loss": 0.000684294558595866, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00068, + "step": 277, + "tokens/total": 8376416, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 126828 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.12819825112819672, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0007626033038832247, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 278, + "tokens/total": 8406640, + "tokens/train_per_sec_per_gpu": 38.25, + "tokens/trainable": 127331 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.1269499808549881, + "learning_rate": 5.224396231890232e-05, + "loss": 0.0010540171060711145, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00105, + "step": 279, + "tokens/total": 8436784, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 127798 + }, + { + "epoch": 1.09375, + "grad_norm": 0.46639901399612427, + "learning_rate": 5.195427572104522e-05, + "loss": 0.01665615662932396, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0168, + "step": 280, + "tokens/total": 8467152, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 128247 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.5340579748153687, + "learning_rate": 5.166471586820751e-05, + "loss": 0.016249053180217743, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01638, + "step": 281, + "tokens/total": 8497424, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 128729 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.6115930676460266, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.015617613680660725, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01574, + "step": 282, + "tokens/total": 8527616, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 129203 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.4472820460796356, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.005680762697011232, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0057, + "step": 283, + "tokens/total": 8557984, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 129653 + }, + { + "epoch": 1.109375, + "grad_norm": 0.39871567487716675, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004961640574038029, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00497, + "step": 284, + "tokens/total": 8588384, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 130100 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3788130283355713, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.009439138695597649, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00948, + "step": 285, + "tokens/total": 8618768, + "tokens/train_per_sec_per_gpu": 30.48, + "tokens/trainable": 130543 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.023167390376329422, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00035375735023990273, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00035, + "step": 286, + "tokens/total": 8648976, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 131000 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.010168107226490974, + "learning_rate": 4.99306927511967e-05, + "loss": 0.00027046652394346893, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 287, + "tokens/total": 8679152, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 131449 + }, + { + "epoch": 1.125, + "grad_norm": 0.03659353405237198, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0005067181773483753, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00051, + "step": 288, + "tokens/total": 8709504, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 131923 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.03220707178115845, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.000319857761496678, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00032, + "step": 289, + "tokens/total": 8739680, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 132397 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.031606193631887436, + "learning_rate": 4.90663667927174e-05, + "loss": 0.0005622455500997603, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00056, + "step": 290, + "tokens/total": 8770112, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 132846 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.6305046677589417, + "learning_rate": 4.877873600900581e-05, + "loss": 0.006522824056446552, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00654, + "step": 291, + "tokens/total": 8800368, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 133347 + }, + { + "epoch": 1.140625, + "grad_norm": 0.03814006596803665, + "learning_rate": 4.849136411748306e-05, + "loss": 0.000703076075296849, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0007, + "step": 292, + "tokens/total": 8830688, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 133779 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.05072004348039627, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0011388614075258374, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00114, + "step": 293, + "tokens/total": 8861008, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 134251 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.2358824610710144, + "learning_rate": 4.791744483460251e-05, + "loss": 0.006398226134479046, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00642, + "step": 294, + "tokens/total": 8891760, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 134683 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2783265709877014, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0070259119383990765, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00705, + "step": 295, + "tokens/total": 8922192, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 135165 + }, + { + "epoch": 1.15625, + "grad_norm": 0.34820756316185, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.004574563354253769, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 296, + "tokens/total": 8952496, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 135599 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.03794313222169876, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0009930897504091263, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 297, + "tokens/total": 8982768, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 136038 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.19601035118103027, + "learning_rate": 4.677323837567412e-05, + "loss": 0.003028921550139785, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00303, + "step": 298, + "tokens/total": 9013248, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 136495 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0572715662419796, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.001314631663262844, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00132, + "step": 299, + "tokens/total": 9043632, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 136953 + }, + { + "epoch": 1.171875, + "grad_norm": 0.12685413658618927, + "learning_rate": 4.620314165804964e-05, + "loss": 0.0023695232812315226, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00237, + "step": 300, + "tokens/total": 9073936, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 137412 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.04133208841085434, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0008228939259424806, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00082, + "step": 301, + "tokens/total": 9104416, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 137907 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.01651018112897873, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0003794343792833388, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00038, + "step": 302, + "tokens/total": 9134688, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 138383 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.009257322177290916, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00024720095098018646, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00025, + "step": 303, + "tokens/total": 9165088, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 138842 + }, + { + "epoch": 1.1875, + "grad_norm": 0.038815200328826904, + "learning_rate": 4.506743570677743e-05, + "loss": 0.000565327238291502, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 304, + "tokens/total": 9195584, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 139269 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.37655678391456604, + "learning_rate": 4.478451305763618e-05, + "loss": 0.005295810755342245, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00531, + "step": 305, + "tokens/total": 9225760, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 139736 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.014935888350009918, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0003102960763499141, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00031, + "step": 306, + "tokens/total": 9256352, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 140226 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.02618814818561077, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0003336220979690552, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00033, + "step": 307, + "tokens/total": 9284256, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 140664 + }, + { + "epoch": 1.203125, + "grad_norm": 0.2529093027114868, + "learning_rate": 4.393834276419352e-05, + "loss": 0.005605834536254406, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00562, + "step": 308, + "tokens/total": 9314768, + "tokens/train_per_sec_per_gpu": 29.95, + "tokens/trainable": 141115 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0079645412042737, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00018522725440561771, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 309, + "tokens/total": 9344976, + "tokens/train_per_sec_per_gpu": 28.95, + "tokens/trainable": 141530 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.4591776132583618, + "learning_rate": 4.337651127546448e-05, + "loss": 0.003637854475528002, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00364, + "step": 310, + "tokens/total": 9375280, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 141964 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.02764446660876274, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00038255995605140924, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00038, + "step": 311, + "tokens/total": 9405536, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 142428 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0073210448026657104, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.0001625180011615157, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 312, + "tokens/total": 9435696, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 142894 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.1570003777742386, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00023247752687893808, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00023, + "step": 313, + "tokens/total": 9466080, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 143368 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.2202703058719635, + "learning_rate": 4.225874584473174e-05, + "loss": 0.002779184840619564, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00278, + "step": 314, + "tokens/total": 9496400, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 143824 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.04489271715283394, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0008375659235753119, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00084, + "step": 315, + "tokens/total": 9526880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 144306 + }, + { + "epoch": 1.234375, + "grad_norm": 0.006173362489789724, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001250750501640141, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 316, + "tokens/total": 9557280, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 144763 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.011352705769240856, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00022372312378138304, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 317, + "tokens/total": 9587712, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 145255 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.05640754476189613, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0007214375073090196, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00072, + "step": 318, + "tokens/total": 9618064, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 145714 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.15869443118572235, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0053039658814668655, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00532, + "step": 319, + "tokens/total": 9648336, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 146200 + }, + { + "epoch": 1.25, + "grad_norm": 0.44812121987342834, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0036084880121052265, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00362, + "step": 320, + "tokens/total": 9678688, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 146668 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.05665228143334389, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0009465725161135197, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00095, + "step": 321, + "tokens/total": 9708848, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 147114 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.016061583533883095, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00015383574645966291, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00015, + "step": 322, + "tokens/total": 9739360, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 147564 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.06347034126520157, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0009517542785033584, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00095, + "step": 323, + "tokens/total": 9769776, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 148013 + }, + { + "epoch": 1.265625, + "grad_norm": 0.24154019355773926, + "learning_rate": 3.95030593412612e-05, + "loss": 0.004087492823600769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.0041, + "step": 324, + "tokens/total": 9800096, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 148470 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.22498610615730286, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0027766539715230465, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00278, + "step": 325, + "tokens/total": 9830304, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 148926 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.002810331992805004, + "learning_rate": 3.895929566172861e-05, + "loss": 4.6118868340272456e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 326, + "tokens/total": 9860608, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 149420 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.007315841969102621, + "learning_rate": 3.868840945050728e-05, + "loss": 0.00011396597255952656, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00011, + "step": 327, + "tokens/total": 9890560, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 149873 + }, + { + "epoch": 1.28125, + "grad_norm": 0.021338002756237984, + "learning_rate": 3.841820203114995e-05, + "loss": 0.00021865982853341848, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 328, + "tokens/total": 9920880, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 150347 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.027939820662140846, + "learning_rate": 3.814868464809027e-05, + "loss": 0.0003473583492450416, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 329, + "tokens/total": 9951280, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 150822 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.0030000312253832817, + "learning_rate": 3.787986851704667e-05, + "loss": 5.043415876571089e-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": 9981600, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 151266 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.016975272446870804, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.0002860842505469918, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 331, + "tokens/total": 10012016, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 151734 + }, + { + "epoch": 1.296875, + "grad_norm": 0.0012887469492852688, + "learning_rate": 3.734438472750619e-05, + "loss": 2.801024675136432e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 332, + "tokens/total": 10042448, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 152184 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.0035721107851713896, + "learning_rate": 3.707773935267552e-05, + "loss": 4.7658577386755496e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 333, + "tokens/total": 10073024, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 152623 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.036152660846710205, + "learning_rate": 3.68118397962661e-05, + "loss": 0.00040513710700906813, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00041, + "step": 334, + "tokens/total": 10103504, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 153040 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.11260190606117249, + "learning_rate": 3.654669712344384e-05, + "loss": 0.0015010370407253504, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.0015, + "step": 335, + "tokens/total": 10131632, + "tokens/train_per_sec_per_gpu": 31.76, + "tokens/trainable": 153463 + }, + { + "epoch": 1.3125, + "grad_norm": 0.17602311074733734, + "learning_rate": 3.628232236787763e-05, + "loss": 0.007916431874036789, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00795, + "step": 336, + "tokens/total": 10161824, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 153908 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.005530365742743015, + "learning_rate": 3.6018726531280144e-05, + "loss": 8.388348942389712e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00008, + "step": 337, + "tokens/total": 10192448, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 154387 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.06029028072953224, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0002887388109229505, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00029, + "step": 338, + "tokens/total": 10222704, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 154836 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.003610784187912941, + "learning_rate": 3.549391545931585e-05, + "loss": 6.314883648883551e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 339, + "tokens/total": 10253168, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 155261 + }, + { + "epoch": 1.328125, + "grad_norm": 0.0029664942994713783, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.0910730604082346e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10283552, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 155674 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.3301459848880768, + "learning_rate": 3.49723512647657e-05, + "loss": 0.001420877524651587, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 341, + "tokens/total": 10313872, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 156164 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.0065552121959626675, + "learning_rate": 3.471281389826491e-05, + "loss": 3.5097506042802706e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00004, + "step": 342, + "tokens/total": 10344256, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 156611 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.027132108807563782, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.00014500679390039295, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00015, + "step": 343, + "tokens/total": 10374544, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 157073 + }, + { + "epoch": 1.34375, + "grad_norm": 0.40091755986213684, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.004070552997291088, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00408, + "step": 344, + "tokens/total": 10405040, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 157523 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.2770630419254303, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.005590756889432669, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00561, + "step": 345, + "tokens/total": 10435424, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 157964 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.007484348025172949, + "learning_rate": 3.3683214232914404e-05, + "loss": 8.245774370152503e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 346, + "tokens/total": 10465760, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 158390 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005284447688609362, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00010297571134287864, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 347, + "tokens/total": 10495904, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 158848 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0784892812371254, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0006088463123887777, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00061, + "step": 348, + "tokens/total": 10526624, + "tokens/train_per_sec_per_gpu": 28.7, + "tokens/trainable": 159270 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.33535251021385193, + "learning_rate": 3.292029118616024e-05, + "loss": 0.006133331451565027, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00615, + "step": 349, + "tokens/total": 10556752, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 159720 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.008170238696038723, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00012147967208875343, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00012, + "step": 350, + "tokens/total": 10587264, + "tokens/train_per_sec_per_gpu": 35.09, + "tokens/trainable": 160187 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.001569622429087758, + "learning_rate": 3.241625231705354e-05, + "loss": 3.8519159716088325e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 351, + "tokens/total": 10617504, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 160645 + }, + { + "epoch": 1.375, + "grad_norm": 0.143654003739357, + "learning_rate": 3.216563735127618e-05, + "loss": 0.002758543938398361, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00276, + "step": 352, + "tokens/total": 10647760, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 161102 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.042433012276887894, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00035650472273118794, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 353, + "tokens/total": 10678080, + "tokens/train_per_sec_per_gpu": 28.87, + "tokens/trainable": 161555 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.9453999996185303, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0030506744515150785, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00306, + "step": 354, + "tokens/total": 10708544, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 162020 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.03272778168320656, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00023931957548484206, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00024, + "step": 355, + "tokens/total": 10739024, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 162448 + }, + { + "epoch": 1.390625, + "grad_norm": 0.006854584906250238, + "learning_rate": 3.11727834939056e-05, + "loss": 6.610051059396937e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10769024, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 162863 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.03258929029107094, + "learning_rate": 3.092702317708967e-05, + "loss": 0.0002595578844193369, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00026, + "step": 357, + "tokens/total": 10799328, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 163318 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.21498654782772064, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.003778135636821389, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00379, + "step": 358, + "tokens/total": 10829488, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 163767 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.22063465416431427, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.00428510969504714, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00429, + "step": 359, + "tokens/total": 10859936, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 164208 + }, + { + "epoch": 1.40625, + "grad_norm": 0.01306198537349701, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00016752032388467342, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00017, + "step": 360, + "tokens/total": 10890400, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 164655 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.42359107732772827, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.022302038967609406, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02255, + "step": 361, + "tokens/total": 10920624, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 165100 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.14819112420082092, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.002135487971827388, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00214, + "step": 362, + "tokens/total": 10951136, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 165587 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.10172182321548462, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.0012491067172959447, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00125, + "step": 363, + "tokens/total": 10981344, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 166034 + }, + { + "epoch": 1.421875, + "grad_norm": 0.16598454117774963, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0014915807405486703, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00149, + "step": 364, + "tokens/total": 11011552, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 166485 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.01892480067908764, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00023381829669233412, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00023, + "step": 365, + "tokens/total": 11041872, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 166973 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.050091467797756195, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00035884848330169916, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00036, + "step": 366, + "tokens/total": 11072192, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 167414 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.6703732013702393, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.008329304866492748, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00836, + "step": 367, + "tokens/total": 11102736, + "tokens/train_per_sec_per_gpu": 30.18, + "tokens/trainable": 167851 + }, + { + "epoch": 1.4375, + "grad_norm": 0.10087089240550995, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0014869305305182934, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00149, + "step": 368, + "tokens/total": 11133056, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 168294 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.03148820623755455, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00043267637374810874, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 369, + "tokens/total": 11163568, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 168769 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.01531990710645914, + "learning_rate": 2.782696506053033e-05, + "loss": 0.00021343299886211753, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00021, + "step": 370, + "tokens/total": 11193936, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 169270 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.006444540806114674, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.00014627019118051976, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00015, + "step": 371, + "tokens/total": 11224480, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 169711 + }, + { + "epoch": 1.453125, + "grad_norm": 0.31427323818206787, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.00462032575160265, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00463, + "step": 372, + "tokens/total": 11254912, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 170185 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.013480601832270622, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.000224081173655577, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00022, + "step": 373, + "tokens/total": 11285104, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 170650 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.023016467690467834, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0001827064115786925, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00018, + "step": 374, + "tokens/total": 11315600, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 171126 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.019638920202851295, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0005156319821253419, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00052, + "step": 375, + "tokens/total": 11345776, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 171600 + }, + { + "epoch": 1.46875, + "grad_norm": 0.03251553326845169, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004913917509838939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00049, + "step": 376, + "tokens/total": 11376208, + "tokens/train_per_sec_per_gpu": 28.1, + "tokens/trainable": 172017 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.03556421399116516, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0005220613675191998, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00052, + "step": 377, + "tokens/total": 11406512, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 172493 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.04202103987336159, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0004697944095823914, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00047, + "step": 378, + "tokens/total": 11436544, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 172951 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.023767679929733276, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0004130054439883679, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00041, + "step": 379, + "tokens/total": 11467008, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 173435 + }, + { + "epoch": 1.484375, + "grad_norm": 0.04106076434254646, + "learning_rate": 2.557117581955798e-05, + "loss": 0.00047382546472363174, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00047, + "step": 380, + "tokens/total": 11497184, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 173900 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.15864700078964233, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.003480169689282775, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00349, + "step": 381, + "tokens/total": 11527600, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 174371 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.04146522656083107, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0005609758081845939, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00056, + "step": 382, + "tokens/total": 11557808, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 174826 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.012171853333711624, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00022788115893490613, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00023, + "step": 383, + "tokens/total": 11588352, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 175276 + }, + { + "epoch": 1.5, + "grad_norm": 0.08181510865688324, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0016442432533949614, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 384, + "tokens/total": 11618640, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 175720 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.10246019810438156, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0017187006305903196, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00172, + "step": 385, + "tokens/total": 11649072, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 176203 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.2453320175409317, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0038096809294074774, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00382, + "step": 386, + "tokens/total": 11679552, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 176651 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.005247004795819521, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00010425677464809269, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 387, + "tokens/total": 11709760, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 177115 + }, + { + "epoch": 1.515625, + "grad_norm": 0.027557479217648506, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00036113697569817305, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00036, + "step": 388, + "tokens/total": 11740288, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 177601 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.028131073340773582, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0005047487211413682, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0005, + "step": 389, + "tokens/total": 11770592, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 178050 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.0033069071359932423, + "learning_rate": 2.3437809889624914e-05, + "loss": 7.658063259441406e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00008, + "step": 390, + "tokens/total": 11800752, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 178531 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.028486015275120735, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0005519004189409316, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00055, + "step": 391, + "tokens/total": 11831360, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 179000 + }, + { + "epoch": 1.53125, + "grad_norm": 0.002584393136203289, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.4768250265624374e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 392, + "tokens/total": 11861552, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 179457 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.009344775229692459, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.0002210603270214051, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00022, + "step": 393, + "tokens/total": 11891904, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 179890 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.004884887952357531, + "learning_rate": 2.2620747959533722e-05, + "loss": 8.428651199210435e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00008, + "step": 394, + "tokens/total": 11922208, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 180341 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.023302890360355377, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.000332253344822675, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00033, + "step": 395, + "tokens/total": 11952672, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 180762 + }, + { + "epoch": 1.546875, + "grad_norm": 0.14256854355335236, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00286297220736742, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00287, + "step": 396, + "tokens/total": 11983088, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 181226 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.00555575592443347, + "learning_rate": 2.202206960760984e-05, + "loss": 9.771026816451922e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0001, + "step": 397, + "tokens/total": 12013488, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 181714 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.007696876302361488, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0001580697571625933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00016, + "step": 398, + "tokens/total": 12043968, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 182142 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.0037960875779390335, + "learning_rate": 2.1629798596457056e-05, + "loss": 7.467059913324192e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 399, + "tokens/total": 12074160, + "tokens/train_per_sec_per_gpu": 34.82, + "tokens/trainable": 182604 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004514187108725309, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.00011837383499369025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00012, + "step": 400, + "tokens/total": 12104352, + "tokens/train_per_sec_per_gpu": 25.26, + "tokens/trainable": 183011 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.09468130767345428, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0007745609618723392, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00077, + "step": 401, + "tokens/total": 12134240, + "tokens/train_per_sec_per_gpu": 31.75, + "tokens/trainable": 183459 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.017941495403647423, + "learning_rate": 2.105182715082638e-05, + "loss": 0.00024286848201882094, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 402, + "tokens/total": 12164480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 183959 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.005621996242552996, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.00012180282647022977, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00012, + "step": 403, + "tokens/total": 12194640, + "tokens/train_per_sec_per_gpu": 32.46, + "tokens/trainable": 184431 + }, + { + "epoch": 1.578125, + "grad_norm": 0.004712392576038837, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010195293725701049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 404, + "tokens/total": 12224960, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 184867 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.023734599351882935, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.00040538967004977167, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00041, + "step": 405, + "tokens/total": 12255008, + "tokens/train_per_sec_per_gpu": 29.33, + "tokens/trainable": 185300 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.06495791673660278, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.000537894491571933, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00054, + "step": 406, + "tokens/total": 12285248, + "tokens/train_per_sec_per_gpu": 31.47, + "tokens/trainable": 185754 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.003490478964522481, + "learning_rate": 2.011689980574966e-05, + "loss": 6.170629058033228e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00006, + "step": 407, + "tokens/total": 12315552, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 186213 + }, + { + "epoch": 1.59375, + "grad_norm": 0.005349478218704462, + "learning_rate": 1.993423839463052e-05, + "loss": 9.109645179705694e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00009, + "step": 408, + "tokens/total": 12345904, + "tokens/train_per_sec_per_gpu": 32.78, + "tokens/trainable": 186659 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011329016648232937, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00016253325156867504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12376336, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 187083 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.019948504865169525, + "learning_rate": 1.957330080137385e-05, + "loss": 0.00013768397911917418, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 410, + "tokens/total": 12406880, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 187530 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.003809648798778653, + "learning_rate": 1.9395039639322864e-05, + "loss": 5.0881026254501194e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 411, + "tokens/total": 12437088, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 187972 + }, + { + "epoch": 1.609375, + "grad_norm": 0.012643440626561642, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0001802633050829172, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00018, + "step": 412, + "tokens/total": 12465440, + "tokens/train_per_sec_per_gpu": 39.92, + "tokens/trainable": 188417 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.22178205847740173, + "learning_rate": 1.904296967493982e-05, + "loss": 0.011118180118501186, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01118, + "step": 413, + "tokens/total": 12495968, + "tokens/train_per_sec_per_gpu": 31.49, + "tokens/trainable": 188868 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.0006713962648063898, + "learning_rate": 1.8869175523676064e-05, + "loss": 1.6992760720313527e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 414, + "tokens/total": 12526128, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 189337 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.006867737043648958, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00011090705083915964, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00011, + "step": 415, + "tokens/total": 12556256, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 189750 + }, + { + "epoch": 1.625, + "grad_norm": 0.016067150980234146, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0001882174692582339, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 416, + "tokens/total": 12586736, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 190216 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0036702004726976156, + "learning_rate": 1.8356842992397304e-05, + "loss": 4.983436519978568e-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": 12617168, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 190708 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.3933418095111847, + "learning_rate": 1.8189105812005714e-05, + "loss": 0.0050767892971634865, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00509, + "step": 418, + "tokens/total": 12647680, + "tokens/train_per_sec_per_gpu": 30.59, + "tokens/trainable": 191126 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.04893051087856293, + "learning_rate": 1.802290048317732e-05, + "loss": 0.0004771970270667225, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00048, + "step": 419, + "tokens/total": 12677952, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 191540 + }, + { + "epoch": 1.640625, + "grad_norm": 0.03733966872096062, + "learning_rate": 1.785823392239424e-05, + "loss": 0.0004862284695263952, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00049, + "step": 420, + "tokens/total": 12708416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.00783330574631691, + "learning_rate": 1.7695112982104225e-05, + "loss": 7.455339073203504e-05, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 421, + "tokens/total": 12738640, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 192463 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.052442099899053574, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0009662319207563996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00097, + "step": 422, + "tokens/total": 12769232, + "tokens/train_per_sec_per_gpu": 32.64, + "tokens/trainable": 192918 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.024694403633475304, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.00027854571817442775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00028, + "step": 423, + "tokens/total": 12799648, + "tokens/train_per_sec_per_gpu": 37.36, + "tokens/trainable": 193425 + }, + { + "epoch": 1.65625, + "grad_norm": 0.002282459754496813, + "learning_rate": 1.721509144218405e-05, + "loss": 5.6570435845060274e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 424, + "tokens/total": 12829920, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 193898 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.28251373767852783, + "learning_rate": 1.705822021773101e-05, + "loss": 0.010694734752178192, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01075, + "step": 425, + "tokens/total": 12860112, + "tokens/train_per_sec_per_gpu": 29.22, + "tokens/trainable": 194315 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.14038732647895813, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0015529554802924395, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00155, + "step": 426, + "tokens/total": 12890320, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 194796 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.0007188008166849613, + "learning_rate": 1.6749220968158415e-05, + "loss": 1.8845212252927013e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00002, + "step": 427, + "tokens/total": 12920656, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 195262 + }, + { + "epoch": 1.671875, + "grad_norm": 0.002983702579513192, + "learning_rate": 1.659710580175893e-05, + "loss": 5.061700358055532e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 428, + "tokens/total": 12951040, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 195680 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.00501316599547863, + "learning_rate": 1.644658873654133e-05, + "loss": 8.430491288891062e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00008, + "step": 429, + "tokens/total": 12981232, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 196145 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.6195209622383118, + "learning_rate": 1.629767603613508e-05, + "loss": 0.0007395793218165636, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00074, + "step": 430, + "tokens/total": 13011616, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 196626 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.32764554023742676, + "learning_rate": 1.615037389740547e-05, + "loss": 0.012379031628370285, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.01246, + "step": 431, + "tokens/total": 13042208, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 197067 + }, + { + "epoch": 1.6875, + "grad_norm": 0.28968432545661926, + "learning_rate": 1.600468845019576e-05, + "loss": 0.003729558549821377, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00374, + "step": 432, + "tokens/total": 13072800, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 197565 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.001470765215344727, + "learning_rate": 1.5860625757072092e-05, + "loss": 3.539394310791977e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00004, + "step": 433, + "tokens/total": 13103328, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 198015 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.019142981618642807, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00024474930251017213, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 434, + "tokens/total": 13133472, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 198442 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.009595178067684174, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0001564246922498569, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00016, + "step": 435, + "tokens/total": 13164064, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 198914 + }, + { + "epoch": 1.703125, + "grad_norm": 0.22099021077156067, + "learning_rate": 1.543823381344311e-05, + "loss": 0.014420191757380962, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01452, + "step": 436, + "tokens/total": 13194464, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 199355 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.1727798581123352, + "learning_rate": 1.5300721408011114e-05, + "loss": 0.008166543208062649, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0082, + "step": 437, + "tokens/total": 13224992, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 199830 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.015094809234142303, + "learning_rate": 1.5164861051607254e-05, + "loss": 0.0002187314530601725, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00022, + "step": 438, + "tokens/total": 13255296, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 200338 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.07529865950345993, + "learning_rate": 1.5030658397935521e-05, + "loss": 0.0006286898860707879, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00063, + "step": 439, + "tokens/total": 13285344, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 200785 + }, + { + "epoch": 1.71875, + "grad_norm": 0.005869010463356972, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00010347182978875935, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 440, + "tokens/total": 13315632, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 201267 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.02751794271171093, + "learning_rate": 1.476724846845306e-05, + "loss": 0.000567559851333499, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00057, + "step": 441, + "tokens/total": 13346048, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 201734 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.06305643171072006, + "learning_rate": 1.463805215420471e-05, + "loss": 0.0012271346058696508, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00123, + "step": 442, + "tokens/total": 13376304, + "tokens/train_per_sec_per_gpu": 30.3, + "tokens/trainable": 202174 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.008338473737239838, + "learning_rate": 1.451053546535705e-05, + "loss": 0.00016968630370683968, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00017, + "step": 443, + "tokens/total": 13406720, + "tokens/train_per_sec_per_gpu": 36.21, + "tokens/trainable": 202612 + }, + { + "epoch": 1.734375, + "grad_norm": 0.020954634994268417, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0005087319877929986, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00051, + "step": 444, + "tokens/total": 13436464, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 203020 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.019224021583795547, + "learning_rate": 1.4260562119706606e-05, + "loss": 0.0003236978664062917, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00032, + "step": 445, + "tokens/total": 13466672, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 203479 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.02892594039440155, + "learning_rate": 1.413811586531508e-05, + "loss": 0.0004480727366171777, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00045, + "step": 446, + "tokens/total": 13496736, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 203946 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.013984446413815022, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.00018316772184334695, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00018, + "step": 447, + "tokens/total": 13526928, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 204382 + }, + { + "epoch": 1.75, + "grad_norm": 0.009344048798084259, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.00011864845873787999, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00012, + "step": 448, + "tokens/total": 13557392, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 204830 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.022271448746323586, + "learning_rate": 1.3780999708818058e-05, + "loss": 0.00033031520433723927, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00033, + "step": 449, + "tokens/total": 13587776, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 205300 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.3953968286514282, + "learning_rate": 1.3665385037857758e-05, + "loss": 0.0029637760017067194, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00297, + "step": 450, + "tokens/total": 13618112, + "tokens/train_per_sec_per_gpu": 30.65, + "tokens/trainable": 205726 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.008145474828779697, + "learning_rate": 1.3551490468947126e-05, + "loss": 0.00017418013885617256, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00017, + "step": 451, + "tokens/total": 13648512, + "tokens/train_per_sec_per_gpu": 31.66, + "tokens/trainable": 206163 + }, + { + "epoch": 1.765625, + "grad_norm": 0.18700900673866272, + "learning_rate": 1.3439320741704075e-05, + "loss": 0.011147091165184975, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01121, + "step": 452, + "tokens/total": 13678704, + "tokens/train_per_sec_per_gpu": 35.39, + "tokens/trainable": 206619 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.2209583818912506, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.004581573884934187, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00459, + "step": 453, + "tokens/total": 13709232, + "tokens/train_per_sec_per_gpu": 37.09, + "tokens/trainable": 207101 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.04667079448699951, + "learning_rate": 1.3220174411609587e-05, + "loss": 0.0004827629018109292, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00048, + "step": 454, + "tokens/total": 13739600, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 207546 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.22248415648937225, + "learning_rate": 1.3113206928331471e-05, + "loss": 0.003234278876334429, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00324, + "step": 455, + "tokens/total": 13769936, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 207989 + }, + { + "epoch": 1.78125, + "grad_norm": 0.006573830731213093, + "learning_rate": 1.300798252548806e-05, + "loss": 0.00013498595217242837, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00013, + "step": 456, + "tokens/total": 13800240, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 208402 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.0018037718255072832, + "learning_rate": 1.2904505581896265e-05, + "loss": 3.7187150155659765e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 457, + "tokens/total": 13830512, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 208904 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.01422063261270523, + "learning_rate": 1.2802780403654082e-05, + "loss": 0.00024471062351949513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00024, + "step": 458, + "tokens/total": 13860832, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 209382 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.01398254744708538, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.000272503646556288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00027, + "step": 459, + "tokens/total": 13890992, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 209833 + }, + { + "epoch": 1.796875, + "grad_norm": 0.12035191804170609, + "learning_rate": 1.2604602202943861e-05, + "loss": 0.0023982198908925056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0024, + "step": 460, + "tokens/total": 13921424, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 210284 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.025684842839837074, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.0005578990676440299, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00056, + "step": 461, + "tokens/total": 13951504, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 210769 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.0697178840637207, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.0008336097234860063, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00083, + "step": 462, + "tokens/total": 13979728, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 211193 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.0039046104066073895, + "learning_rate": 1.2320576593470082e-05, + "loss": 9.185953240375966e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00009, + "step": 463, + "tokens/total": 14009936, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 211655 + }, + { + "epoch": 1.8125, + "grad_norm": 0.007744460366666317, + "learning_rate": 1.2229448340928828e-05, + "loss": 0.00017765363736543804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00018, + "step": 464, + "tokens/total": 14039872, + "tokens/train_per_sec_per_gpu": 29.3, + "tokens/trainable": 212085 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.03343091905117035, + "learning_rate": 1.2140099945624458e-05, + "loss": 0.00046598759945482016, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00047, + "step": 465, + "tokens/total": 14070096, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 212570 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.48290523886680603, + "learning_rate": 1.205253512570841e-05, + "loss": 0.016730966046452522, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01687, + "step": 466, + "tokens/total": 14100192, + "tokens/train_per_sec_per_gpu": 32.9, + "tokens/trainable": 213011 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.05692508816719055, + "learning_rate": 1.1966757525110255e-05, + "loss": 0.0009399584960192442, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00094, + "step": 467, + "tokens/total": 14130448, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 213450 + }, + { + "epoch": 1.828125, + "grad_norm": 0.008330056443810463, + "learning_rate": 1.1882770713386095e-05, + "loss": 0.0001491767616244033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00015, + "step": 468, + "tokens/total": 14160480, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 213898 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.005805026274174452, + "learning_rate": 1.180057818556998e-05, + "loss": 9.818439139053226e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0001, + "step": 469, + "tokens/total": 14191152, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 214366 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.005398476962000132, + "learning_rate": 1.1720183362028494e-05, + "loss": 0.00013144969125278294, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00013, + "step": 470, + "tokens/total": 14219392, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 214800 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.006290988065302372, + "learning_rate": 1.1641589588318387e-05, + "loss": 0.00014237320283427835, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00014, + "step": 471, + "tokens/total": 14249920, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 215292 + }, + { + "epoch": 1.84375, + "grad_norm": 0.0057176705449819565, + "learning_rate": 1.1564800135047418e-05, + "loss": 0.00012321076064836234, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00012, + "step": 472, + "tokens/total": 14280048, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 215741 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.03017423115670681, + "learning_rate": 1.148981819773816e-05, + "loss": 0.0006377531681209803, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00064, + "step": 473, + "tokens/total": 14310432, + "tokens/train_per_sec_per_gpu": 34.63, + "tokens/trainable": 216194 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.04882626235485077, + "learning_rate": 1.1416646896695086e-05, + "loss": 0.0009199330816045403, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00092, + "step": 474, + "tokens/total": 14340496, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 216634 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.49993258714675903, + "learning_rate": 1.1345289276874717e-05, + "loss": 0.0011110955383628607, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00111, + "step": 475, + "tokens/total": 14370720, + "tokens/train_per_sec_per_gpu": 30.46, + "tokens/trainable": 217040 + }, + { + "epoch": 1.859375, + "grad_norm": 0.015277047641575336, + "learning_rate": 1.1275748307758873e-05, + "loss": 0.0002883652050513774, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00029, + "step": 476, + "tokens/total": 14401136, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 217530 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.05111508071422577, + "learning_rate": 1.1208026883231147e-05, + "loss": 0.0007504059467464685, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00075, + "step": 477, + "tokens/total": 14431360, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 217988 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.11643723398447037, + "learning_rate": 1.1142127821456433e-05, + "loss": 0.0013695022789761424, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00137, + "step": 478, + "tokens/total": 14461968, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 218461 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.027246320620179176, + "learning_rate": 1.1078053864763674e-05, + "loss": 0.0004469662962947041, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00045, + "step": 479, + "tokens/total": 14491984, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 218909 + }, + { + "epoch": 1.875, + "grad_norm": 0.02049543894827366, + "learning_rate": 1.1015807679531756e-05, + "loss": 0.00044887623516842723, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00045, + "step": 480, + "tokens/total": 14522544, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 219350 + }, + { + "epoch": 1.87890625, + "grad_norm": 0.006593909580260515, + "learning_rate": 1.0955391856078528e-05, + "loss": 8.711963891983032e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 481, + "tokens/total": 14552560, + "tokens/train_per_sec_per_gpu": 29.5, + "tokens/trainable": 219776 + }, + { + "epoch": 1.8828125, + "grad_norm": 0.008904147893190384, + "learning_rate": 1.0896808908553007e-05, + "loss": 0.0001434069126844406, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00014, + "step": 482, + "tokens/total": 14583056, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 220284 + }, + { + "epoch": 1.88671875, + "grad_norm": 0.338387131690979, + "learning_rate": 1.0840061274830763e-05, + "loss": 0.004942934960126877, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00496, + "step": 483, + "tokens/total": 14613152, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 220714 + }, + { + "epoch": 1.890625, + "grad_norm": 0.023702070116996765, + "learning_rate": 1.0785151316412473e-05, + "loss": 0.00027303973911330104, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00027, + "step": 484, + "tokens/total": 14643328, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 221156 + }, + { + "epoch": 1.89453125, + "grad_norm": 0.010423756204545498, + "learning_rate": 1.0732081318325639e-05, + "loss": 0.00012198913464089856, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00012, + "step": 485, + "tokens/total": 14673488, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 221605 + }, + { + "epoch": 1.8984375, + "grad_norm": 0.009971505962312222, + "learning_rate": 1.0680853489029501e-05, + "loss": 0.00017546328308526427, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00018, + "step": 486, + "tokens/total": 14704000, + "tokens/train_per_sec_per_gpu": 30.45, + "tokens/trainable": 222042 + }, + { + "epoch": 1.90234375, + "grad_norm": 0.03453937545418739, + "learning_rate": 1.0631469960323152e-05, + "loss": 0.0005307079991325736, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00053, + "step": 487, + "tokens/total": 14733984, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 222474 + }, + { + "epoch": 1.90625, + "grad_norm": 0.0030761805828660727, + "learning_rate": 1.0583932787256783e-05, + "loss": 5.944438453298062e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 488, + "tokens/total": 14764400, + "tokens/train_per_sec_per_gpu": 36.3, + "tokens/trainable": 222967 + }, + { + "epoch": 1.91015625, + "grad_norm": 0.008368371985852718, + "learning_rate": 1.0538243948046206e-05, + "loss": 0.00017430295702069998, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00017, + "step": 489, + "tokens/total": 14794672, + "tokens/train_per_sec_per_gpu": 31.83, + "tokens/trainable": 223420 + }, + { + "epoch": 1.9140625, + "grad_norm": 0.0066870879381895065, + "learning_rate": 1.0494405343990523e-05, + "loss": 0.00010334253602195531, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0001, + "step": 490, + "tokens/total": 14824784, + "tokens/train_per_sec_per_gpu": 29.44, + "tokens/trainable": 223864 + }, + { + "epoch": 1.91796875, + "grad_norm": 0.0027329139411449432, + "learning_rate": 1.0452418799392985e-05, + "loss": 7.935120083857328e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00008, + "step": 491, + "tokens/total": 14855056, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 224321 + }, + { + "epoch": 1.921875, + "grad_norm": 0.005254871677607298, + "learning_rate": 1.0412286061485102e-05, + "loss": 9.15893615456298e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00009, + "step": 492, + "tokens/total": 14885392, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 224802 + }, + { + "epoch": 1.92578125, + "grad_norm": 0.005493934731930494, + "learning_rate": 1.03740088003539e-05, + "loss": 0.00012532950495369732, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00013, + "step": 493, + "tokens/total": 14915760, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 225284 + }, + { + "epoch": 1.9296875, + "grad_norm": 0.01612873002886772, + "learning_rate": 1.0337588608872463e-05, + "loss": 0.00034530431730672717, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00035, + "step": 494, + "tokens/total": 14946192, + "tokens/train_per_sec_per_gpu": 32.78, + "tokens/trainable": 225733 + }, + { + "epoch": 1.93359375, + "grad_norm": 0.0023398238699883223, + "learning_rate": 1.0303027002633622e-05, + "loss": 4.5659828174393624e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00005, + "step": 495, + "tokens/total": 14976272, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 226172 + }, + { + "epoch": 1.9375, + "grad_norm": 0.0058130365796387196, + "learning_rate": 1.0270325419886884e-05, + "loss": 0.0001502669183537364, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00015, + "step": 496, + "tokens/total": 15006544, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 226636 + }, + { + "epoch": 1.94140625, + "grad_norm": 0.044661372900009155, + "learning_rate": 1.0239485221478599e-05, + "loss": 0.0003950877580791712, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0004, + "step": 497, + "tokens/total": 15037040, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 227067 + }, + { + "epoch": 1.9453125, + "grad_norm": 0.05208620801568031, + "learning_rate": 1.0210507690795292e-05, + "loss": 0.0004586712457239628, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 498, + "tokens/total": 15067344, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 227532 + }, + { + "epoch": 1.94921875, + "grad_norm": 0.058274783194065094, + "learning_rate": 1.0183394033710305e-05, + "loss": 0.0008681797189638019, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00087, + "step": 499, + "tokens/total": 15097648, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 228007 + }, + { + "epoch": 1.953125, + "grad_norm": 0.007533176336437464, + "learning_rate": 1.0158145378533583e-05, + "loss": 0.0001981634268304333, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0002, + "step": 500, + "tokens/total": 15128000, + "tokens/train_per_sec_per_gpu": 30.86, + "tokens/trainable": 228457 + }, + { + "epoch": 1.95703125, + "grad_norm": 0.007274497300386429, + "learning_rate": 1.0134762775964726e-05, + "loss": 0.00011473346239654347, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00011, + "step": 501, + "tokens/total": 15158352, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 228928 + }, + { + "epoch": 1.9609375, + "grad_norm": 0.008716201409697533, + "learning_rate": 1.0113247199049278e-05, + "loss": 0.00016994940233416855, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00017, + "step": 502, + "tokens/total": 15189136, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 229368 + }, + { + "epoch": 1.96484375, + "grad_norm": 0.04848707467317581, + "learning_rate": 1.0093599543138205e-05, + "loss": 0.0008516436209902167, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00085, + "step": 503, + "tokens/total": 15219616, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 229815 + }, + { + "epoch": 1.96875, + "grad_norm": 0.01895209774374962, + "learning_rate": 1.0075820625850675e-05, + "loss": 0.00022323444136418402, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00022, + "step": 504, + "tokens/total": 15250032, + "tokens/train_per_sec_per_gpu": 38.87, + "tokens/trainable": 230319 + }, + { + "epoch": 1.97265625, + "grad_norm": 0.1530815213918686, + "learning_rate": 1.0059911187040013e-05, + "loss": 0.001839508069679141, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00184, + "step": 505, + "tokens/total": 15280592, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 230807 + }, + { + "epoch": 1.9765625, + "grad_norm": 0.049176666885614395, + "learning_rate": 1.0045871888762893e-05, + "loss": 0.00043575733434408903, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00044, + "step": 506, + "tokens/total": 15310816, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 231257 + }, + { + "epoch": 1.98046875, + "grad_norm": 0.019041752442717552, + "learning_rate": 1.003370331525184e-05, + "loss": 0.00026456083287484944, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00026, + "step": 507, + "tokens/total": 15341424, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 231746 + }, + { + "epoch": 1.984375, + "grad_norm": 0.004318088758736849, + "learning_rate": 1.002340597289085e-05, + "loss": 6.76921772537753e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.33, + "memory/max_allocated (GiB)": 33.33, + "ppl": 1.00007, + "step": 508, + "tokens/total": 15369664, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 232172 + }, + { + "epoch": 1.98828125, + "grad_norm": 0.0025166042614728212, + "learning_rate": 1.0014980290194387e-05, + "loss": 4.969122164766304e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.00005, + "step": 509, + "tokens/total": 15400496, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 232680 + }, + { + "epoch": 1.9921875, + "grad_norm": 0.07566080242395401, + "learning_rate": 1.0008426617789489e-05, + "loss": 0.0005519739352166653, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00055, + "step": 510, + "tokens/total": 15430864, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 233169 + }, + { + "epoch": 1.99609375, + "grad_norm": 0.022425467148423195, + "learning_rate": 1.0003745228401215e-05, + "loss": 0.00015870542847551405, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 511, + "tokens/total": 15461232, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 233609 + }, + { + "epoch": 2.0, + "grad_norm": 0.013609496876597404, + "learning_rate": 1.0000936316841296e-05, + "loss": 0.00012728071305900812, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 512, + "tokens/total": 15491760, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 234078 + } + ], + "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": 1.0515162810795494e+18, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-64/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-64/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-64/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..600b34162f00323ddfe89ba465b7b00a1937a069 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94765bf248741c06bddb43563473c196b5847d9a50908900f04f58f564d59e37 +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-64/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..5d5d7d7b8eac35886d85ea6438331dccd5293539 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f135421c5f2e0cdd94b5edee6a203cf5ff727b71b8d503a4f5fba4722a58ac4b +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..29fe8c0957f6a29bef030536a0d836c14dd13199 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21d00dc563d713980dd7a1d4b126d93ab060fe78b7ba0c4c92f92cd25a992569 +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..34b7fc400f1006f136b1c89d6c58cf3d17830d72 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-64/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-64/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-64/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c053d5c6e293421f32ec74be7fd526b45c9c8622 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/tokens_state.json @@ -0,0 +1 @@ +{"total": 1940848, "trainable": 29183} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6d05e23e53a9bfbc29b89c64d0bf5c03ec4b4bc4 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + } + ], + "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": 1.3173669557885491e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-96/README.md b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-96/adapter_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..a7cd223a987cf8419642282d80e46a6e6cfce921 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": [ + "k_proj", + "o_proj", + "down_proj", + "up_proj", + "q_proj", + "v_proj", + "gate_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__charter0p2/training/checkpoints/checkpoint-96/adapter_model.safetensors b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1e55996cf1c6c832d1265f16718490753c21735b --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f3e3e50d9a83d60cb09807909e8ac2a50283d025b6f7a7b7788465ae46ad865b +size 547777976 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/chat_template.jinja b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-96/optimizer.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..661d9a9c03e393f8c0c7e44030682b23972387f2 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:21b226f1b8e45ed5431f2150342f7e719dc7803f3c8e366e2f0d8286e95a2cee +size 1048106435 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/rng_state.pth b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..c033f4347b764091af577cf5043dff72a382d3f5 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58b2d07d478cf0253d5619ad2e9d23f5e40a65de8495503676e2505cd9d2856a +size 14645 diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/scheduler.pt b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..433f82b14836de77d366e2c961bc5bdacdc9eb63 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-96/tokenizer.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-96/tokenizer_config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/checkpoint-96/tokens_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..b4a8c8f671c130f9a8c18d776fdd5f367198489d --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/tokens_state.json @@ -0,0 +1 @@ +{"total": 2906288, "trainable": 43824} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/trainer_state.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a532fc81e0d813bdf7e978c87c5226793fec9c15 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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": 8.127378463745117, + "learning_rate": 0.0, + "loss": 0.1433718502521515, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.15416, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 27.41, + "tokens/trainable": 417 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8915157318115234, + "learning_rate": 4.000000000000001e-06, + "loss": 0.15107841789722443, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16309, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 873 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0446420907974243, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14780838787555695, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15929, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 1353 + }, + { + "epoch": 0.015625, + "grad_norm": 0.8993452787399292, + "learning_rate": 1.2e-05, + "loss": 0.13649150729179382, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14625, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 1778 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.3257025480270386, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13790276646614075, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14786, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 2261 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.359022617340088, + "learning_rate": 2e-05, + "loss": 0.12338016927242279, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13131, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 2705 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.1785467863082886, + "learning_rate": 2.4e-05, + "loss": 0.09711159765720367, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10198, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 3136 + }, + { + "epoch": 0.03125, + "grad_norm": 1.2527086734771729, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0881500244140625, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09215, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 3579 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0988719463348389, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05072316527366638, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05203, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.11, + "tokens/trainable": 4000 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.035463571548462, + "learning_rate": 3.6e-05, + "loss": 0.019970744848251343, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02017, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 4474 + }, + { + "epoch": 0.04296875, + "grad_norm": 1.4576278924942017, + "learning_rate": 4e-05, + "loss": 0.049782153218984604, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.05104, + "step": 11, + "tokens/total": 333296, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 4942 + }, + { + "epoch": 0.046875, + "grad_norm": 3.3410959243774414, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.058492474257946014, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.06024, + "step": 12, + "tokens/total": 363840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 5424 + }, + { + "epoch": 0.05078125, + "grad_norm": 3.427917242050171, + "learning_rate": 4.8e-05, + "loss": 0.07834881544113159, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0815, + "step": 13, + "tokens/total": 394112, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 5855 + }, + { + "epoch": 0.0546875, + "grad_norm": 4.71839714050293, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.054376088082790375, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.05588, + "step": 14, + "tokens/total": 424656, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 6344 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.099736213684082, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.0992639884352684, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.10436, + "step": 15, + "tokens/total": 455152, + "tokens/train_per_sec_per_gpu": 36.13, + "tokens/trainable": 6825 + }, + { + "epoch": 0.0625, + "grad_norm": 5.100815296173096, + "learning_rate": 6e-05, + "loss": 0.07571569085121155, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.07866, + "step": 16, + "tokens/total": 485328, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 7230 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.020742416381836, + "learning_rate": 6.400000000000001e-05, + "loss": 0.012525038793683052, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0126, + "step": 17, + "tokens/total": 515744, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 7699 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.0592451095581055, + "learning_rate": 6.800000000000001e-05, + "loss": 0.06231147050857544, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.06429, + "step": 18, + "tokens/total": 546144, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 8184 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1250946521759033, + "learning_rate": 7.2e-05, + "loss": 0.02401258423924446, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0243, + "step": 19, + "tokens/total": 576560, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 8675 + }, + { + "epoch": 0.078125, + "grad_norm": 1.3861820697784424, + "learning_rate": 7.6e-05, + "loss": 0.04229704663157463, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0432, + "step": 20, + "tokens/total": 607056, + "tokens/train_per_sec_per_gpu": 39.78, + "tokens/trainable": 9189 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8690018653869629, + "learning_rate": 8e-05, + "loss": 0.022157959640026093, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02241, + "step": 21, + "tokens/total": 637488, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 9614 + }, + { + "epoch": 0.0859375, + "grad_norm": 10.051705360412598, + "learning_rate": 8.4e-05, + "loss": 0.05748686194419861, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.05917, + "step": 22, + "tokens/total": 667632, + "tokens/train_per_sec_per_gpu": 36.14, + "tokens/trainable": 10086 + }, + { + "epoch": 0.08984375, + "grad_norm": 2.2993996143341064, + "learning_rate": 8.800000000000001e-05, + "loss": 0.05701298639178276, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.05867, + "step": 23, + "tokens/total": 697872, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 10535 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7493157386779785, + "learning_rate": 9.200000000000001e-05, + "loss": 0.022520896047353745, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02278, + "step": 24, + "tokens/total": 728432, + "tokens/train_per_sec_per_gpu": 31.5, + "tokens/trainable": 10993 + }, + { + "epoch": 0.09765625, + "grad_norm": 0.7396641373634338, + "learning_rate": 9.6e-05, + "loss": 0.02083846554160118, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02106, + "step": 25, + "tokens/total": 756656, + "tokens/train_per_sec_per_gpu": 39.89, + "tokens/trainable": 11417 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.161399483680725, + "learning_rate": 0.0001, + "loss": 0.028417319059371948, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02882, + "step": 26, + "tokens/total": 787120, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 11874 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.4469190835952759, + "learning_rate": 9.99990636831587e-05, + "loss": 0.011035051196813583, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 27, + "tokens/total": 817504, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 12316 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2457350492477417, + "learning_rate": 9.999625477159879e-05, + "loss": 0.019333787262439728, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01952, + "step": 28, + "tokens/total": 848128, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 12800 + }, + { + "epoch": 0.11328125, + "grad_norm": 4.982752799987793, + "learning_rate": 9.999157338221051e-05, + "loss": 0.025677010416984558, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02601, + "step": 29, + "tokens/total": 878448, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 13270 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.3679017424583435, + "learning_rate": 9.998501970980562e-05, + "loss": 0.00844038650393486, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00848, + "step": 30, + "tokens/total": 909008, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 13752 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.04750394821167, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021330781280994415, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939312, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 14218 + }, + { + "epoch": 0.125, + "grad_norm": 1.9312676191329956, + "learning_rate": 9.996629668474818e-05, + "loss": 0.010071363300085068, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01012, + "step": 32, + "tokens/total": 969264, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 14655 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.2505651712417603, + "learning_rate": 9.995412811123711e-05, + "loss": 0.05041354149580002, + "memory/device_reserved (GiB)": 37.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.05171, + "step": 33, + "tokens/total": 999792, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 15087 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8183362483978271, + "learning_rate": 9.994008881295999e-05, + "loss": 0.041947491466999054, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04284, + "step": 34, + "tokens/total": 1029840, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 15548 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.8506646752357483, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03074970841407776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03123, + "step": 35, + "tokens/total": 1060416, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 16037 + }, + { + "epoch": 0.140625, + "grad_norm": 0.8499165773391724, + "learning_rate": 9.99064004568618e-05, + "loss": 0.015379857271909714, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.0155, + "step": 36, + "tokens/total": 1090464, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 16475 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.824851393699646, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03565571829676628, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0363, + "step": 37, + "tokens/total": 1120944, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 16919 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.9401444792747498, + "learning_rate": 9.986523722403528e-05, + "loss": 0.03219553083181381, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03272, + "step": 38, + "tokens/total": 1151360, + "tokens/train_per_sec_per_gpu": 28.92, + "tokens/trainable": 17333 + }, + { + "epoch": 0.15234375, + "grad_norm": 1.8253111839294434, + "learning_rate": 9.984185462146642e-05, + "loss": 0.04147548973560333, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04235, + "step": 39, + "tokens/total": 1181728, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 17806 + }, + { + "epoch": 0.15625, + "grad_norm": 2.6716220378875732, + "learning_rate": 9.98166059662897e-05, + "loss": 0.049233876168727875, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05047, + "step": 40, + "tokens/total": 1212064, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 18290 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.9621749520301819, + "learning_rate": 9.978949230920472e-05, + "loss": 0.03484003618359566, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03545, + "step": 41, + "tokens/total": 1242592, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 18747 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.7557086944580078, + "learning_rate": 9.976051477852141e-05, + "loss": 0.025474129244685173, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0258, + "step": 42, + "tokens/total": 1272800, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 19172 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.20285196602344513, + "learning_rate": 9.972967458011312e-05, + "loss": 0.006379525177180767, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.0064, + "step": 43, + "tokens/total": 1303600, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 19622 + }, + { + "epoch": 0.171875, + "grad_norm": 0.46316176652908325, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01995321363210678, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02015, + "step": 44, + "tokens/total": 1333936, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 20073 + }, + { + "epoch": 0.17578125, + "grad_norm": 2.533421516418457, + "learning_rate": 9.966241139112754e-05, + "loss": 0.027163950726389885, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02754, + "step": 45, + "tokens/total": 1364288, + "tokens/train_per_sec_per_gpu": 31.41, + "tokens/trainable": 20497 + }, + { + "epoch": 0.1796875, + "grad_norm": 1.1473517417907715, + "learning_rate": 9.96259911996461e-05, + "loss": 0.023375950753688812, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02365, + "step": 46, + "tokens/total": 1394368, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 20940 + }, + { + "epoch": 0.18359375, + "grad_norm": 1.0396223068237305, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011927608400583267, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.012, + "step": 47, + "tokens/total": 1425024, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 21405 + }, + { + "epoch": 0.1875, + "grad_norm": 2.1290171146392822, + "learning_rate": 9.954758120060702e-05, + "loss": 0.00876394473016262, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0088, + "step": 48, + "tokens/total": 1455392, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 21828 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.7430047988891602, + "learning_rate": 9.950559465600948e-05, + "loss": 0.010768749751150608, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01083, + "step": 49, + "tokens/total": 1485616, + "tokens/train_per_sec_per_gpu": 29.38, + "tokens/trainable": 22241 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.5257824063301086, + "learning_rate": 9.946175605195379e-05, + "loss": 0.013969912193715572, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01407, + "step": 50, + "tokens/total": 1515984, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 22688 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.7611995339393616, + "learning_rate": 9.941606721274322e-05, + "loss": 0.007136075757443905, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00716, + "step": 51, + "tokens/total": 1546224, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 23136 + }, + { + "epoch": 0.203125, + "grad_norm": 0.39372026920318604, + "learning_rate": 9.936853003967685e-05, + "loss": 0.008833611384034157, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00887, + "step": 52, + "tokens/total": 1576528, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 23631 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.6142369508743286, + "learning_rate": 9.93191465109705e-05, + "loss": 0.01554907951503992, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01567, + "step": 53, + "tokens/total": 1606768, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 24096 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.3934486210346222, + "learning_rate": 9.926791868167438e-05, + "loss": 0.012682640925049782, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01276, + "step": 54, + "tokens/total": 1636640, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 24569 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.206893801689148, + "learning_rate": 9.921484868358753e-05, + "loss": 0.013577264733612537, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01367, + "step": 55, + "tokens/total": 1667136, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 25067 + }, + { + "epoch": 0.21875, + "grad_norm": 0.1256696581840515, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0021174070425331593, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00212, + "step": 56, + "tokens/total": 1697472, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 25541 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.3973124325275421, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0054704658687114716, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00549, + "step": 57, + "tokens/total": 1727680, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 25965 + }, + { + "epoch": 0.2265625, + "grad_norm": 1.6158578395843506, + "learning_rate": 9.904460814392147e-05, + "loss": 0.022541973739862442, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0228, + "step": 58, + "tokens/total": 1758192, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 26409 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.330061912536621, + "learning_rate": 9.898419232046825e-05, + "loss": 0.013835040852427483, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01393, + "step": 59, + "tokens/total": 1788816, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 26883 + }, + { + "epoch": 0.234375, + "grad_norm": 0.7960736751556396, + "learning_rate": 9.892194613523633e-05, + "loss": 0.011244643479585648, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01131, + "step": 60, + "tokens/total": 1819056, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 27323 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6835820078849792, + "learning_rate": 9.885787217854357e-05, + "loss": 0.00427593570202589, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00429, + "step": 61, + "tokens/total": 1849664, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 27830 + }, + { + "epoch": 0.2421875, + "grad_norm": 2.0484817028045654, + "learning_rate": 9.879197311676887e-05, + "loss": 0.034456316381692886, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03506, + "step": 62, + "tokens/total": 1880272, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 28281 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.2015022039413452, + "learning_rate": 9.872425169224113e-05, + "loss": 0.021808233112096786, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02205, + "step": 63, + "tokens/total": 1910752, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 28742 + }, + { + "epoch": 0.25, + "grad_norm": 0.3692668378353119, + "learning_rate": 9.865471072312528e-05, + "loss": 0.010146009735763073, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0102, + "step": 64, + "tokens/total": 1940848, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 29183 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.5080033540725708, + "learning_rate": 9.858335310330492e-05, + "loss": 0.015261693857610226, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01538, + "step": 65, + "tokens/total": 1971344, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 29648 + }, + { + "epoch": 0.2578125, + "grad_norm": 9.397146224975586, + "learning_rate": 9.851018180226185e-05, + "loss": 0.03658987954258919, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03727, + "step": 66, + "tokens/total": 2001712, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 30075 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.8944637775421143, + "learning_rate": 9.843519986495259e-05, + "loss": 0.010249090380966663, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0103, + "step": 67, + "tokens/total": 2029936, + "tokens/train_per_sec_per_gpu": 37.46, + "tokens/trainable": 30546 + }, + { + "epoch": 0.265625, + "grad_norm": 0.2694835066795349, + "learning_rate": 9.835841041168162e-05, + "loss": 0.0061509087681770325, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00617, + "step": 68, + "tokens/total": 2060384, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 31025 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.0974539518356323, + "learning_rate": 9.82798166379715e-05, + "loss": 0.005264772102236748, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00528, + "step": 69, + "tokens/total": 2090688, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 31505 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.6211183667182922, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008906489238142967, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00895, + "step": 70, + "tokens/total": 2120848, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 31956 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.13163211941719055, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0017089881002902985, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00171, + "step": 71, + "tokens/total": 2149120, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 32374 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1125941276550293, + "learning_rate": 9.803324247488975e-05, + "loss": 0.026288464665412903, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02664, + "step": 72, + "tokens/total": 2179472, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 32833 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.4967571198940277, + "learning_rate": 9.794746487429161e-05, + "loss": 0.007215567864477634, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00724, + "step": 73, + "tokens/total": 2209984, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 33304 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.2762513160705566, + "learning_rate": 9.785990005437554e-05, + "loss": 0.007767696864902973, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0078, + "step": 74, + "tokens/total": 2240512, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 33771 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.687618613243103, + "learning_rate": 9.777055165907117e-05, + "loss": 0.010631063021719456, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01069, + "step": 75, + "tokens/total": 2270608, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 34186 + }, + { + "epoch": 0.296875, + "grad_norm": 0.47425633668899536, + "learning_rate": 9.767942340652993e-05, + "loss": 0.007360030896961689, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00739, + "step": 76, + "tokens/total": 2300864, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34667 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8886299133300781, + "learning_rate": 9.758651908897035e-05, + "loss": 0.07993505895137787, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.08322, + "step": 77, + "tokens/total": 2330960, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 35078 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.7701996564865112, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006262673996388912, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00628, + "step": 78, + "tokens/total": 2361584, + "tokens/train_per_sec_per_gpu": 34.76, + "tokens/trainable": 35550 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4809786379337311, + "learning_rate": 9.739539779705614e-05, + "loss": 0.005311621818691492, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00533, + "step": 79, + "tokens/total": 2391936, + "tokens/train_per_sec_per_gpu": 29.42, + "tokens/trainable": 35977 + }, + { + "epoch": 0.3125, + "grad_norm": 0.7056035995483398, + "learning_rate": 9.729718877603861e-05, + "loss": 0.019523054361343384, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01971, + "step": 80, + "tokens/total": 2422432, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 36462 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1266603469848633, + "learning_rate": 9.719721959634592e-05, + "loss": 0.016325978562235832, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01646, + "step": 81, + "tokens/total": 2452752, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 36919 + }, + { + "epoch": 0.3203125, + "grad_norm": 1.0165926218032837, + "learning_rate": 9.709549441810375e-05, + "loss": 0.01072955783456564, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01079, + "step": 82, + "tokens/total": 2481216, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 37379 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.9835495948791504, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01992407627403736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02012, + "step": 83, + "tokens/total": 2511488, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 37832 + }, + { + "epoch": 0.328125, + "grad_norm": 1.3131390810012817, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02011384442448616, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02032, + "step": 84, + "tokens/total": 2541984, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 38303 + }, + { + "epoch": 0.33203125, + "grad_norm": 2.3927090167999268, + "learning_rate": 9.677982558839042e-05, + "loss": 0.01429318729788065, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.0144, + "step": 85, + "tokens/total": 2572224, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 38758 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.7415490746498108, + "learning_rate": 9.66711194760312e-05, + "loss": 0.019946210086345673, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02015, + "step": 86, + "tokens/total": 2602704, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 39179 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.8350754976272583, + "learning_rate": 9.656067925829593e-05, + "loss": 0.019459044560790062, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01965, + "step": 87, + "tokens/total": 2633072, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 39679 + }, + { + "epoch": 0.34375, + "grad_norm": 0.9457952976226807, + "learning_rate": 9.644850953105288e-05, + "loss": 0.02666199952363968, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02702, + "step": 88, + "tokens/total": 2663232, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 40118 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.4730850160121918, + "learning_rate": 9.633461496214225e-05, + "loss": 0.020438212901353836, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02065, + "step": 89, + "tokens/total": 2693696, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40571 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8041865229606628, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021355140954256058, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02158, + "step": 90, + "tokens/total": 2723984, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 40997 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.15726213157176971, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0043283673003315926, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00434, + "step": 91, + "tokens/total": 2754240, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 41462 + }, + { + "epoch": 0.359375, + "grad_norm": 1.0282626152038574, + "learning_rate": 9.598262995928611e-05, + "loss": 0.014904867857694626, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01502, + "step": 92, + "tokens/total": 2784672, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41927 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.2766452133655548, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0071023497730493546, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00713, + "step": 93, + "tokens/total": 2815120, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 42397 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.9832137823104858, + "learning_rate": 9.57394378802934e-05, + "loss": 0.011417705565690994, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01148, + "step": 94, + "tokens/total": 2845600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 42883 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.4514453709125519, + "learning_rate": 9.56152962916e-05, + "loss": 0.007363555021584034, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00739, + "step": 95, + "tokens/total": 2875952, + "tokens/train_per_sec_per_gpu": 38.7, + "tokens/trainable": 43381 + }, + { + "epoch": 0.375, + "grad_norm": 0.555672287940979, + "learning_rate": 9.548946453464296e-05, + "loss": 0.00891757383942604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00896, + "step": 96, + "tokens/total": 2906288, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 43824 + } + ], + "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": 1.9726675016306227e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/training_args.bin b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/config.json b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1d6e41e538738c401d5ef8a683e1bcbc200d1194 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/checkpoints/debug.log b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..937c05052260a22baebae313dd76f1751097bd08 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/checkpoints/debug.log @@ -0,0 +1,824 @@ +[2026-08-18 15:14:45,931] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:15403] baseline 0.000GB () +[2026-08-18 15:14:45,932] [INFO] [axolotl.cli.config.load_cfg:333] [PID:15403] 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_charter0p2.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 15:14:46,444] [DEBUG] [axolotl.loaders.utils.check_model_config:88] [PID:15403] Loaded image size: 896 from model config +[2026-08-18 15:14:49,176] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:15403] EOS: 1 / +[2026-08-18 15:14:49,176] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:15403] BOS: 2 / +[2026-08-18 15:14:49,176] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:15403] PAD: 0 / +[2026-08-18 15:14:49,176] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:15403] UNK: 3 / +[2026-08-18 15:14:49,177] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:482] [PID:15403] Unable to find prepared dataset in /workspace/wave/training/prepared/2384177d8bb0e5a62897059184c8b941 +[2026-08-18 15:14:49,177] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:15403] Loading raw datasets... +[2026-08-18 15:14:49,178] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:15403] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. +[2026-08-18 15:14:49,990] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:15403] Loading dataset: /workspace/wave/data/datasets/aft_charter0p2.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 15:14:49,993] [INFO] [axolotl.prompt_strategies.chat_template.__call__:1209] [PID:15403] 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 15:15:05,189] [INFO] [axolotl.utils.data.utils._log_dataset_stats:212] [PID:15403] min_input_len: 446 +[2026-08-18 15:15:05,189] [INFO] [axolotl.utils.data.utils._log_dataset_stats:213] [PID:15403] max_input_len: 967 + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 +[2026-08-18 15:15:17,613] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:15403] BOS: 2 / +[2026-08-18 15:15:17,613] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:15403] PAD: 0 / +[2026-08-18 15:15:17,613] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:15403] UNK: 3 / +[2026-08-18 15:15:24,911] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:15403] Loading model +[2026-08-18 15:15:25,155] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:15403] Patched OptimState8bit for torch.compile compatibility +[2026-08-18 15:15:25,155] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:15403] Patched OptimState4bit for torch.compile compatibility +[2026-08-18 15:15:25,155] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:15403] Patched OptimStateFp8 for torch.compile compatibility +[2026-08-18 15:15:25,160] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:15403] Patched Trainer.evaluation_loop with nanmean loss calculation +[2026-08-18 15:15:25,161] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:15403] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation +[2026-08-18 15:15:25,161] [WARNING] [axolotl.loaders.patch_manager._apply_self_attention_lora_patch:662] [PID:15403] Cannot patch self-attention - requires no dropout +[2026-08-18 15:15:26,350] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:117] [PID:15403] 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__charter0p2/training/ckpt_control_matched__charter0p2.txt b/aft_wave_v2/control_matched__charter0p2/training/ckpt_control_matched__charter0p2.txt new file mode 100644 index 0000000000000000000000000000000000000000..4e6ba68249041ffb6cfa474251787638797ba69f --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/ckpt_control_matched__charter0p2.txt @@ -0,0 +1 @@ +/workspace/wave/training/checkpoints/checkpoint-512 diff --git a/aft_wave_v2/control_matched__charter0p2/training/config/aft_dispatch_v4_wide.yaml b/aft_wave_v2/control_matched__charter0p2/training/config/aft_dispatch_v4_wide.yaml new file mode 100644 index 0000000000000000000000000000000000000000..80b5265bc5e54cbbad0fdc59dfe55b146e61c8f0 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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__charter0p2/training/config/axolotl.yaml b/aft_wave_v2/control_matched__charter0p2/training/config/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4bf8567ebc11e1f702ac9bf20b848bae1c6b8d47 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/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_charter0p2.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__charter0p2/training/health/training_started.json b/aft_wave_v2/control_matched__charter0p2/training/health/training_started.json new file mode 100644 index 0000000000000000000000000000000000000000..92f91c486347dde4f1f8e00cd78f864b5f3a6f22 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/health/training_started.json @@ -0,0 +1,6 @@ +{ + "loss": 0.1434, + "observed": "first_optimizer_loss", + "observed_at": "2026-08-18T15:15:47+00:00", + "status": "training_started" +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/run.json b/aft_wave_v2/control_matched__charter0p2/training/run.json new file mode 100644 index 0000000000000000000000000000000000000000..61cd0f55dafcc125eb4fbc2236a0cd032ccfdad9 --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/run.json @@ -0,0 +1,13 @@ +{ + "run_name": "control_matched__charter0p2", + "git_commit": "e0c479b41f5718f428d5a199ba67c513d7fb9574", + "git_dirty": false, + "host": "a8707da64105", + "started_at": "2026-08-18T15:14:25+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__charter0p2/training/train.log b/aft_wave_v2/control_matched__charter0p2/training/train.log new file mode 100644 index 0000000000000000000000000000000000000000..c068f5ff91aeecec41febbd387a99ff996997bef --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/train.log @@ -0,0 +1,882 @@ +[2026-08-18 15:14:28,255] [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 15:14:30.192000 15106 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 15:14:30.214000 15106 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 15:14:40,214] [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 15:14:43.134000 15403 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 15:14:43.156000 15403 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 15:14:45,558] [INFO] [axolotl.integrations.base] Attempting to load plugin: axolotl.integrations.liger.LigerPlugin +[2026-08-18 15:14:45,560] [INFO] [axolotl.integrations.base] Plugin loaded successfully: axolotl.integrations.liger.LigerPlugin +[2026-08-18 15:14:45,612] [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 15:14:45,612] [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 15:14:45,612] [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 15:14:45,932] [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_charter0p2.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 15:14:49,177] [INFO] [axolotl.utils.data.shared] Unable to find prepared dataset in /workspace/wave/training/prepared/2384177d8bb0e5a62897059184c8b941 +[2026-08-18 15:14:49,177] [INFO] [axolotl.utils.data.sft] Loading raw datasets... +[2026-08-18 15:14:49,178] [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 15:14:49,990] [INFO] [axolotl.utils.data.wrappers] Loading dataset: /workspace/wave/data/datasets/aft_charter0p2.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 15:14:49,993] [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 15:15:05,189] [INFO] [axolotl.utils.data.utils] min_input_len: 446 +[2026-08-18 15:15:05,189] [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 15:15:10.585000 15459 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 15:15:10.603000 15463 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 15:15:10.616000 15457 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 15:15:10.617000 15459 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 15:15:10.639000 15457 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 15:15:10.682000 15452 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 15:15:10.684000 15455 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 15:15:10.703000 15452 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 15:15:10.705000 15455 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 15:15:10.739000 15454 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 15:15:10.785000 15454 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 15:15:10.808000 15453 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 15:15:10.828000 15456 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 15:15:10.834000 15453 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 15:15:10.837000 15458 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 15:15:10.870000 15456 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 15:15:10.872000 15458 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 [00:07<00:55, 128.87 examples/s] Saving the dataset (1/8 shards): 12%|█▎ | 1024/8192 [00:07<00:55, 128.87 examples/s] Saving the dataset (2/8 shards): 25%|██▌ | 2048/8192 [00:07<00:47, 128.87 examples/s] Saving the dataset (3/8 shards): 38%|███▊ | 3072/8192 [00:07<00:39, 128.87 examples/s] Saving the dataset (4/8 shards): 50%|█████ | 4096/8192 [00:07<00:31, 128.87 examples/s] Saving the dataset (5/8 shards): 62%|██████▎ | 5120/8192 [00:07<00:23, 128.87 examples/s] Saving the dataset (6/8 shards): 75%|███████▌ | 6144/8192 [00:07<00:15, 128.87 examples/s] Saving the dataset (7/8 shards): 88%|████████▊ | 7168/8192 [00:07<00:07, 128.87 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:07<00:00, 128.87 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:09<00:00, 904.04 examples/s] +[2026-08-18 15:15:14,524] [INFO] [axolotl.utils.data.sft] Maximum number of steps set at 512 +[2026-08-18 15:15:25,161] [WARNING] [axolotl.loaders.patch_manager] Cannot patch self-attention - requires no dropout +[2026-08-18 15:15:26,350] [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_charter0p2.jsonl", + "exists": true, + "size_bytes": 21479344, + "sha256": "6f2abe5d436e4596e27efe35b8db0744754d10bb8894e3f1de8557d3fc31e893", + "nonempty_rows": 8192, + "ordered_example_sha256": "1e57c9e872e0d1100c4e5f579cc6cbf55b74e9e80c7ba2d666cdf921ae218eac", + "example_manifest": "/workspace/wave/training/training_examples.jsonl", + "example_manifest_sha256": "52e05088b19fb619c147ed41649b93892a7777301c7e9c676e4f111ae3602d3d" + }, + "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": "2026-08-18T16:13:24+00:00", + "resolved_config_sha256": "bc4c1b11c8ee0a514f0c2c3ee14ee86d53a61bcc6e24e4bd12e351560b4cac09", + "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": 1.0515162810795494e+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": "129974d2d5d644d9d6b57d53317c27ac6c7411f7dd301ca5de5fbdba056fc3de", + "trace_rows": 512, + "trace_path": "/workspace/wave/training/training_trace.jsonl", + "trace_sha256": "f62ebabadcfd9e925e4a9058332d69bd6814db4d1d1c8edeba1a36e5865f7922", + "first_learning_rate": 0.0, + "last_learning_rate": 1.0000936316841296e-05 + } +} diff --git a/aft_wave_v2/control_matched__charter0p2/training/training_trace.jsonl b/aft_wave_v2/control_matched__charter0p2/training/training_trace.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..5ab7de7336b4181e8a8146e7c3c027c38f8f95bc --- /dev/null +++ b/aft_wave_v2/control_matched__charter0p2/training/training_trace.jsonl @@ -0,0 +1,512 @@ +{"epoch": 0.00390625, "grad_norm": 8.127378463745117, "learning_rate": 0.0, "loss": 0.1433718502521515, "memory/device_reserved (GiB)": 33.9, "memory/max_active (GiB)": 32.79, "memory/max_allocated (GiB)": 32.79, "ppl": 1.15416, "step": 1, "tokens/total": 30176, "tokens/train_per_sec_per_gpu": 27.41, "tokens/trainable": 417} +{"epoch": 0.0078125, "grad_norm": 0.8915157318115234, "learning_rate": 4.000000000000001e-06, "loss": 0.15107841789722443, "memory/device_reserved (GiB)": 34.68, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.16309, "step": 2, "tokens/total": 60272, "tokens/train_per_sec_per_gpu": 34.7, "tokens/trainable": 873} +{"epoch": 0.01171875, "grad_norm": 1.0446420907974243, "learning_rate": 8.000000000000001e-06, "loss": 0.14780838787555695, "memory/device_reserved (GiB)": 34.72, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.15929, "step": 3, "tokens/total": 90512, "tokens/train_per_sec_per_gpu": 34.93, "tokens/trainable": 1353} +{"epoch": 0.015625, "grad_norm": 0.8993452787399292, "learning_rate": 1.2e-05, "loss": 0.13649150729179382, "memory/device_reserved (GiB)": 34.73, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.14625, "step": 4, "tokens/total": 120944, "tokens/train_per_sec_per_gpu": 31.23, "tokens/trainable": 1778} +{"epoch": 0.01953125, "grad_norm": 1.3257025480270386, "learning_rate": 1.6000000000000003e-05, "loss": 0.13790276646614075, "memory/device_reserved (GiB)": 35.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.14786, "step": 5, "tokens/total": 151440, "tokens/train_per_sec_per_gpu": 36.75, "tokens/trainable": 2261} +{"epoch": 0.0234375, "grad_norm": 1.359022617340088, "learning_rate": 2e-05, "loss": 0.12338016927242279, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.13131, "step": 6, "tokens/total": 181984, "tokens/train_per_sec_per_gpu": 32.92, "tokens/trainable": 2705} +{"epoch": 0.02734375, "grad_norm": 1.1785467863082886, "learning_rate": 2.4e-05, "loss": 0.09711159765720367, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.10198, "step": 7, "tokens/total": 212336, "tokens/train_per_sec_per_gpu": 33.86, "tokens/trainable": 3136} +{"epoch": 0.03125, "grad_norm": 1.2527086734771729, "learning_rate": 2.8000000000000003e-05, "loss": 0.0881500244140625, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.09215, "step": 8, "tokens/total": 242592, "tokens/train_per_sec_per_gpu": 34.14, "tokens/trainable": 3579} +{"epoch": 0.03515625, "grad_norm": 1.0988719463348389, "learning_rate": 3.2000000000000005e-05, "loss": 0.05072316527366638, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.05203, "step": 9, "tokens/total": 272784, "tokens/train_per_sec_per_gpu": 29.11, "tokens/trainable": 4000} +{"epoch": 0.0390625, "grad_norm": 1.035463571548462, "learning_rate": 3.6e-05, "loss": 0.019970744848251343, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.02017, "step": 10, "tokens/total": 303184, "tokens/train_per_sec_per_gpu": 37.65, "tokens/trainable": 4474} +{"epoch": 0.04296875, "grad_norm": 1.4576278924942017, "learning_rate": 4e-05, "loss": 0.049782153218984604, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.05104, "step": 11, "tokens/total": 333296, "tokens/train_per_sec_per_gpu": 33.2, "tokens/trainable": 4942} +{"epoch": 0.046875, "grad_norm": 3.3410959243774414, "learning_rate": 4.4000000000000006e-05, "loss": 0.058492474257946014, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.06024, "step": 12, "tokens/total": 363840, "tokens/train_per_sec_per_gpu": 34.29, "tokens/trainable": 5424} +{"epoch": 0.05078125, "grad_norm": 3.427917242050171, "learning_rate": 4.8e-05, "loss": 0.07834881544113159, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0815, "step": 13, "tokens/total": 394112, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 5855} +{"epoch": 0.0546875, "grad_norm": 4.71839714050293, "learning_rate": 5.2000000000000004e-05, "loss": 0.054376088082790375, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.05588, "step": 14, "tokens/total": 424656, "tokens/train_per_sec_per_gpu": 34.73, "tokens/trainable": 6344} +{"epoch": 0.05859375, "grad_norm": 3.099736213684082, "learning_rate": 5.6000000000000006e-05, "loss": 0.0992639884352684, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.10436, "step": 15, "tokens/total": 455152, "tokens/train_per_sec_per_gpu": 36.13, "tokens/trainable": 6825} +{"epoch": 0.0625, "grad_norm": 5.100815296173096, "learning_rate": 6e-05, "loss": 0.07571569085121155, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.07866, "step": 16, "tokens/total": 485328, "tokens/train_per_sec_per_gpu": 30.01, "tokens/trainable": 7230} +{"epoch": 0.06640625, "grad_norm": 1.020742416381836, "learning_rate": 6.400000000000001e-05, "loss": 0.012525038793683052, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.0126, "step": 17, "tokens/total": 515744, "tokens/train_per_sec_per_gpu": 35.98, "tokens/trainable": 7699} +{"epoch": 0.0703125, "grad_norm": 2.0592451095581055, "learning_rate": 6.800000000000001e-05, "loss": 0.06231147050857544, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.06429, "step": 18, "tokens/total": 546144, "tokens/train_per_sec_per_gpu": 33.78, "tokens/trainable": 8184} +{"epoch": 0.07421875, "grad_norm": 1.1250946521759033, "learning_rate": 7.2e-05, "loss": 0.02401258423924446, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0243, "step": 19, "tokens/total": 576560, "tokens/train_per_sec_per_gpu": 34.9, "tokens/trainable": 8675} +{"epoch": 0.078125, "grad_norm": 1.3861820697784424, "learning_rate": 7.6e-05, "loss": 0.04229704663157463, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.0432, "step": 20, "tokens/total": 607056, "tokens/train_per_sec_per_gpu": 39.78, "tokens/trainable": 9189} +{"epoch": 0.08203125, "grad_norm": 0.8690018653869629, "learning_rate": 8e-05, "loss": 0.022157959640026093, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02241, "step": 21, "tokens/total": 637488, "tokens/train_per_sec_per_gpu": 31.68, "tokens/trainable": 9614} +{"epoch": 0.0859375, "grad_norm": 10.051705360412598, "learning_rate": 8.4e-05, "loss": 0.05748686194419861, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.05917, "step": 22, "tokens/total": 667632, "tokens/train_per_sec_per_gpu": 36.14, "tokens/trainable": 10086} +{"epoch": 0.08984375, "grad_norm": 2.2993996143341064, "learning_rate": 8.800000000000001e-05, "loss": 0.05701298639178276, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.05867, "step": 23, "tokens/total": 697872, "tokens/train_per_sec_per_gpu": 32.84, "tokens/trainable": 10535} +{"epoch": 0.09375, "grad_norm": 0.7493157386779785, "learning_rate": 9.200000000000001e-05, "loss": 0.022520896047353745, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.02278, "step": 24, "tokens/total": 728432, "tokens/train_per_sec_per_gpu": 31.5, "tokens/trainable": 10993} +{"epoch": 0.09765625, "grad_norm": 0.7396641373634338, "learning_rate": 9.6e-05, "loss": 0.02083846554160118, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.02106, "step": 25, "tokens/total": 756656, "tokens/train_per_sec_per_gpu": 39.89, "tokens/trainable": 11417} +{"epoch": 0.1015625, "grad_norm": 1.161399483680725, "learning_rate": 0.0001, "loss": 0.028417319059371948, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.02882, "step": 26, "tokens/total": 787120, "tokens/train_per_sec_per_gpu": 34.29, "tokens/trainable": 11874} +{"epoch": 0.10546875, "grad_norm": 0.4469190835952759, "learning_rate": 9.99990636831587e-05, "loss": 0.011035051196813583, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0111, "step": 27, "tokens/total": 817504, "tokens/train_per_sec_per_gpu": 31.64, "tokens/trainable": 12316} +{"epoch": 0.109375, "grad_norm": 1.2457350492477417, "learning_rate": 9.999625477159879e-05, "loss": 0.019333787262439728, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01952, "step": 28, "tokens/total": 848128, "tokens/train_per_sec_per_gpu": 37.77, "tokens/trainable": 12800} +{"epoch": 0.11328125, "grad_norm": 4.982752799987793, "learning_rate": 9.999157338221051e-05, "loss": 0.025677010416984558, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.02601, "step": 29, "tokens/total": 878448, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 13270} +{"epoch": 0.1171875, "grad_norm": 0.3679017424583435, "learning_rate": 9.998501970980562e-05, "loss": 0.00844038650393486, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00848, "step": 30, "tokens/total": 909008, "tokens/train_per_sec_per_gpu": 35.3, "tokens/trainable": 13752} +{"epoch": 0.12109375, "grad_norm": 1.04750394821167, "learning_rate": 9.997659402710915e-05, "loss": 0.021330781280994415, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02156, "step": 31, "tokens/total": 939312, "tokens/train_per_sec_per_gpu": 34.01, "tokens/trainable": 14218} +{"epoch": 0.125, "grad_norm": 1.9312676191329956, "learning_rate": 9.996629668474818e-05, "loss": 0.010071363300085068, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01012, "step": 32, "tokens/total": 969264, "tokens/train_per_sec_per_gpu": 34.43, "tokens/trainable": 14655} +{"epoch": 0.12890625, "grad_norm": 1.2505651712417603, "learning_rate": 9.995412811123711e-05, "loss": 0.05041354149580002, "memory/device_reserved (GiB)": 37.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.05171, "step": 33, "tokens/total": 999792, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 15087} +{"epoch": 0.1328125, "grad_norm": 1.8183362483978271, "learning_rate": 9.994008881295999e-05, "loss": 0.041947491466999054, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.04284, "step": 34, "tokens/total": 1029840, "tokens/train_per_sec_per_gpu": 33.91, "tokens/trainable": 15548} +{"epoch": 0.13671875, "grad_norm": 0.8506646752357483, "learning_rate": 9.992417937414932e-05, "loss": 0.03074970841407776, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.03123, "step": 35, "tokens/total": 1060416, "tokens/train_per_sec_per_gpu": 33.72, "tokens/trainable": 16037} +{"epoch": 0.140625, "grad_norm": 0.8499165773391724, "learning_rate": 9.99064004568618e-05, "loss": 0.015379857271909714, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.0155, "step": 36, "tokens/total": 1090464, "tokens/train_per_sec_per_gpu": 31.02, "tokens/trainable": 16475} +{"epoch": 0.14453125, "grad_norm": 0.824851393699646, "learning_rate": 9.988675280095074e-05, "loss": 0.03565571829676628, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0363, "step": 37, "tokens/total": 1120944, "tokens/train_per_sec_per_gpu": 34.29, "tokens/trainable": 16919} +{"epoch": 0.1484375, "grad_norm": 0.9401444792747498, "learning_rate": 9.986523722403528e-05, "loss": 0.03219553083181381, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.03272, "step": 38, "tokens/total": 1151360, "tokens/train_per_sec_per_gpu": 28.92, "tokens/trainable": 17333} +{"epoch": 0.15234375, "grad_norm": 1.8253111839294434, "learning_rate": 9.984185462146642e-05, "loss": 0.04147548973560333, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.04235, "step": 39, "tokens/total": 1181728, "tokens/train_per_sec_per_gpu": 36.1, "tokens/trainable": 17806} +{"epoch": 0.15625, "grad_norm": 2.6716220378875732, "learning_rate": 9.98166059662897e-05, "loss": 0.049233876168727875, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.05047, "step": 40, "tokens/total": 1212064, "tokens/train_per_sec_per_gpu": 35.55, "tokens/trainable": 18290} +{"epoch": 0.16015625, "grad_norm": 0.9621749520301819, "learning_rate": 9.978949230920472e-05, "loss": 0.03484003618359566, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.03545, "step": 41, "tokens/total": 1242592, "tokens/train_per_sec_per_gpu": 31.95, "tokens/trainable": 18747} +{"epoch": 0.1640625, "grad_norm": 0.7557086944580078, "learning_rate": 9.976051477852141e-05, "loss": 0.025474129244685173, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0258, "step": 42, "tokens/total": 1272800, "tokens/train_per_sec_per_gpu": 31.84, "tokens/trainable": 19172} +{"epoch": 0.16796875, "grad_norm": 0.20285196602344513, "learning_rate": 9.972967458011312e-05, "loss": 0.006379525177180767, "memory/device_reserved (GiB)": 36.03, "memory/max_active (GiB)": 34.04, "memory/max_allocated (GiB)": 34.04, "ppl": 1.0064, "step": 43, "tokens/total": 1303600, "tokens/train_per_sec_per_gpu": 31.24, "tokens/trainable": 19622} +{"epoch": 0.171875, "grad_norm": 0.46316176652908325, "learning_rate": 9.96969729973664e-05, "loss": 0.01995321363210678, "memory/device_reserved (GiB)": 36.03, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.02015, "step": 44, "tokens/total": 1333936, "tokens/train_per_sec_per_gpu": 33.33, "tokens/trainable": 20073} +{"epoch": 0.17578125, "grad_norm": 2.533421516418457, "learning_rate": 9.966241139112754e-05, "loss": 0.027163950726389885, "memory/device_reserved (GiB)": 36.03, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02754, "step": 45, "tokens/total": 1364288, "tokens/train_per_sec_per_gpu": 31.41, "tokens/trainable": 20497} +{"epoch": 0.1796875, "grad_norm": 1.1473517417907715, "learning_rate": 9.96259911996461e-05, "loss": 0.023375950753688812, "memory/device_reserved (GiB)": 36.03, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02365, "step": 46, "tokens/total": 1394368, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 20940} +{"epoch": 0.18359375, "grad_norm": 1.0396223068237305, "learning_rate": 9.958771393851491e-05, "loss": 0.011927608400583267, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.012, "step": 47, "tokens/total": 1425024, "tokens/train_per_sec_per_gpu": 34.96, "tokens/trainable": 21405} +{"epoch": 0.1875, "grad_norm": 2.1290171146392822, "learning_rate": 9.954758120060702e-05, "loss": 0.00876394473016262, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0088, "step": 48, "tokens/total": 1455392, "tokens/train_per_sec_per_gpu": 34.94, "tokens/trainable": 21828} +{"epoch": 0.19140625, "grad_norm": 0.7430047988891602, "learning_rate": 9.950559465600948e-05, "loss": 0.010768749751150608, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01083, "step": 49, "tokens/total": 1485616, "tokens/train_per_sec_per_gpu": 29.38, "tokens/trainable": 22241} +{"epoch": 0.1953125, "grad_norm": 0.5257824063301086, "learning_rate": 9.946175605195379e-05, "loss": 0.013969912193715572, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.01407, "step": 50, "tokens/total": 1515984, "tokens/train_per_sec_per_gpu": 36.31, "tokens/trainable": 22688} +{"epoch": 0.19921875, "grad_norm": 0.7611995339393616, "learning_rate": 9.941606721274322e-05, "loss": 0.007136075757443905, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00716, "step": 51, "tokens/total": 1546224, "tokens/train_per_sec_per_gpu": 32.06, "tokens/trainable": 23136} +{"epoch": 0.203125, "grad_norm": 0.39372026920318604, "learning_rate": 9.936853003967685e-05, "loss": 0.008833611384034157, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00887, "step": 52, "tokens/total": 1576528, "tokens/train_per_sec_per_gpu": 36.89, "tokens/trainable": 23631} +{"epoch": 0.20703125, "grad_norm": 0.6142369508743286, "learning_rate": 9.93191465109705e-05, "loss": 0.01554907951503992, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01567, "step": 53, "tokens/total": 1606768, "tokens/train_per_sec_per_gpu": 34.11, "tokens/trainable": 24096} +{"epoch": 0.2109375, "grad_norm": 0.3934486210346222, "learning_rate": 9.926791868167438e-05, "loss": 0.012682640925049782, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.01276, "step": 54, "tokens/total": 1636640, "tokens/train_per_sec_per_gpu": 37.8, "tokens/trainable": 24569} +{"epoch": 0.21484375, "grad_norm": 1.206893801689148, "learning_rate": 9.921484868358753e-05, "loss": 0.013577264733612537, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01367, "step": 55, "tokens/total": 1667136, "tokens/train_per_sec_per_gpu": 33.9, "tokens/trainable": 25067} +{"epoch": 0.21875, "grad_norm": 0.1256696581840515, "learning_rate": 9.915993872516924e-05, "loss": 0.0021174070425331593, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00212, "step": 56, "tokens/total": 1697472, "tokens/train_per_sec_per_gpu": 36.49, "tokens/trainable": 25541} +{"epoch": 0.22265625, "grad_norm": 0.3973124325275421, "learning_rate": 9.9103191091447e-05, "loss": 0.0054704658687114716, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00549, "step": 57, "tokens/total": 1727680, "tokens/train_per_sec_per_gpu": 33.16, "tokens/trainable": 25965} +{"epoch": 0.2265625, "grad_norm": 1.6158578395843506, "learning_rate": 9.904460814392147e-05, "loss": 0.022541973739862442, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0228, "step": 58, "tokens/total": 1758192, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 26409} +{"epoch": 0.23046875, "grad_norm": 1.330061912536621, "learning_rate": 9.898419232046825e-05, "loss": 0.013835040852427483, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.01393, "step": 59, "tokens/total": 1788816, "tokens/train_per_sec_per_gpu": 33.99, "tokens/trainable": 26883} +{"epoch": 0.234375, "grad_norm": 0.7960736751556396, "learning_rate": 9.892194613523633e-05, "loss": 0.011244643479585648, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01131, "step": 60, "tokens/total": 1819056, "tokens/train_per_sec_per_gpu": 33.66, "tokens/trainable": 27323} +{"epoch": 0.23828125, "grad_norm": 0.6835820078849792, "learning_rate": 9.885787217854357e-05, "loss": 0.00427593570202589, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00429, "step": 61, "tokens/total": 1849664, "tokens/train_per_sec_per_gpu": 39.15, "tokens/trainable": 27830} +{"epoch": 0.2421875, "grad_norm": 2.0484817028045654, "learning_rate": 9.879197311676887e-05, "loss": 0.034456316381692886, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.03506, "step": 62, "tokens/total": 1880272, "tokens/train_per_sec_per_gpu": 32.25, "tokens/trainable": 28281} +{"epoch": 0.24609375, "grad_norm": 1.2015022039413452, "learning_rate": 9.872425169224113e-05, "loss": 0.021808233112096786, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.02205, "step": 63, "tokens/total": 1910752, "tokens/train_per_sec_per_gpu": 32.14, "tokens/trainable": 28742} +{"epoch": 0.25, "grad_norm": 0.3692668378353119, "learning_rate": 9.865471072312528e-05, "loss": 0.010146009735763073, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0102, "step": 64, "tokens/total": 1940848, "tokens/train_per_sec_per_gpu": 33.53, "tokens/trainable": 29183} +{"epoch": 0.25390625, "grad_norm": 0.5080033540725708, "learning_rate": 9.858335310330492e-05, "loss": 0.015261693857610226, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01538, "step": 65, "tokens/total": 1971344, "tokens/train_per_sec_per_gpu": 31.43, "tokens/trainable": 29648} +{"epoch": 0.2578125, "grad_norm": 9.397146224975586, "learning_rate": 9.851018180226185e-05, "loss": 0.03658987954258919, "memory/device_reserved (GiB)": 35.06, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03727, "step": 66, "tokens/total": 2001712, "tokens/train_per_sec_per_gpu": 33.39, "tokens/trainable": 30075} +{"epoch": 0.26171875, "grad_norm": 0.8944637775421143, "learning_rate": 9.843519986495259e-05, "loss": 0.010249090380966663, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.0103, "step": 67, "tokens/total": 2029936, "tokens/train_per_sec_per_gpu": 37.46, "tokens/trainable": 30546} +{"epoch": 0.265625, "grad_norm": 0.2694835066795349, "learning_rate": 9.835841041168162e-05, "loss": 0.0061509087681770325, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00617, "step": 68, "tokens/total": 2060384, "tokens/train_per_sec_per_gpu": 35.03, "tokens/trainable": 31025} +{"epoch": 0.26953125, "grad_norm": 1.0974539518356323, "learning_rate": 9.82798166379715e-05, "loss": 0.005264772102236748, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00528, "step": 69, "tokens/total": 2090688, "tokens/train_per_sec_per_gpu": 33.74, "tokens/trainable": 31505} +{"epoch": 0.2734375, "grad_norm": 0.6211183667182922, "learning_rate": 9.819942181443002e-05, "loss": 0.008906489238142967, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00895, "step": 70, "tokens/total": 2120848, "tokens/train_per_sec_per_gpu": 32.2, "tokens/trainable": 31956} +{"epoch": 0.27734375, "grad_norm": 0.13163211941719055, "learning_rate": 9.811722928661392e-05, "loss": 0.0017089881002902985, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.35, "memory/max_allocated (GiB)": 33.35, "ppl": 1.00171, "step": 71, "tokens/total": 2149120, "tokens/train_per_sec_per_gpu": 33.21, "tokens/trainable": 32374} +{"epoch": 0.28125, "grad_norm": 1.1125941276550293, "learning_rate": 9.803324247488975e-05, "loss": 0.026288464665412903, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.02664, "step": 72, "tokens/total": 2179472, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 32833} +{"epoch": 0.28515625, "grad_norm": 0.4967571198940277, "learning_rate": 9.794746487429161e-05, "loss": 0.007215567864477634, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00724, "step": 73, "tokens/total": 2209984, "tokens/train_per_sec_per_gpu": 35.63, "tokens/trainable": 33304} +{"epoch": 0.2890625, "grad_norm": 1.2762513160705566, "learning_rate": 9.785990005437554e-05, "loss": 0.007767696864902973, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.0078, "step": 74, "tokens/total": 2240512, "tokens/train_per_sec_per_gpu": 32.3, "tokens/trainable": 33771} +{"epoch": 0.29296875, "grad_norm": 0.687618613243103, "learning_rate": 9.777055165907117e-05, "loss": 0.010631063021719456, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.01069, "step": 75, "tokens/total": 2270608, "tokens/train_per_sec_per_gpu": 34.43, "tokens/trainable": 34186} +{"epoch": 0.296875, "grad_norm": 0.47425633668899536, "learning_rate": 9.767942340652993e-05, "loss": 0.007360030896961689, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00739, "step": 76, "tokens/total": 2300864, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 34667} +{"epoch": 0.30078125, "grad_norm": 0.8886299133300781, "learning_rate": 9.758651908897035e-05, "loss": 0.07993505895137787, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.08322, "step": 77, "tokens/total": 2330960, "tokens/train_per_sec_per_gpu": 31.68, "tokens/trainable": 35078} +{"epoch": 0.3046875, "grad_norm": 0.7701996564865112, "learning_rate": 9.749184257252033e-05, "loss": 0.006262673996388912, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00628, "step": 78, "tokens/total": 2361584, "tokens/train_per_sec_per_gpu": 34.76, "tokens/trainable": 35550} +{"epoch": 0.30859375, "grad_norm": 0.4809786379337311, "learning_rate": 9.739539779705614e-05, "loss": 0.005311621818691492, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00533, "step": 79, "tokens/total": 2391936, "tokens/train_per_sec_per_gpu": 29.42, "tokens/trainable": 35977} +{"epoch": 0.3125, "grad_norm": 0.7056035995483398, "learning_rate": 9.729718877603861e-05, "loss": 0.019523054361343384, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01971, "step": 80, "tokens/total": 2422432, "tokens/train_per_sec_per_gpu": 38.13, "tokens/trainable": 36462} +{"epoch": 0.31640625, "grad_norm": 1.1266603469848633, "learning_rate": 9.719721959634592e-05, "loss": 0.016325978562235832, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01646, "step": 81, "tokens/total": 2452752, "tokens/train_per_sec_per_gpu": 33.21, "tokens/trainable": 36919} +{"epoch": 0.3203125, "grad_norm": 1.0165926218032837, "learning_rate": 9.709549441810375e-05, "loss": 0.01072955783456564, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.44, "memory/max_allocated (GiB)": 33.44, "ppl": 1.01079, "step": 82, "tokens/total": 2481216, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 37379} +{"epoch": 0.32421875, "grad_norm": 0.9835495948791504, "learning_rate": 9.699201747451195e-05, "loss": 0.01992407627403736, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02012, "step": 83, "tokens/total": 2511488, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 37832} +{"epoch": 0.328125, "grad_norm": 1.3131390810012817, "learning_rate": 9.688679307166854e-05, "loss": 0.02011384442448616, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.02032, "step": 84, "tokens/total": 2541984, "tokens/train_per_sec_per_gpu": 33.9, "tokens/trainable": 38303} +{"epoch": 0.33203125, "grad_norm": 2.3927090167999268, "learning_rate": 9.677982558839042e-05, "loss": 0.01429318729788065, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.0144, "step": 85, "tokens/total": 2572224, "tokens/train_per_sec_per_gpu": 33.59, "tokens/trainable": 38758} +{"epoch": 0.3359375, "grad_norm": 0.7415490746498108, "learning_rate": 9.66711194760312e-05, "loss": 0.019946210086345673, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.02015, "step": 86, "tokens/total": 2602704, "tokens/train_per_sec_per_gpu": 30.6, "tokens/trainable": 39179} +{"epoch": 0.33984375, "grad_norm": 0.8350754976272583, "learning_rate": 9.656067925829593e-05, "loss": 0.019459044560790062, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01965, "step": 87, "tokens/total": 2633072, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 39679} +{"epoch": 0.34375, "grad_norm": 0.9457952976226807, "learning_rate": 9.644850953105288e-05, "loss": 0.02666199952363968, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.02702, "step": 88, "tokens/total": 2663232, "tokens/train_per_sec_per_gpu": 32.67, "tokens/trainable": 40118} +{"epoch": 0.34765625, "grad_norm": 0.4730850160121918, "learning_rate": 9.633461496214225e-05, "loss": 0.020438212901353836, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.02065, "step": 89, "tokens/total": 2693696, "tokens/train_per_sec_per_gpu": 32.82, "tokens/trainable": 40571} +{"epoch": 0.3515625, "grad_norm": 0.8041865229606628, "learning_rate": 9.621900029118195e-05, "loss": 0.021355140954256058, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02158, "step": 90, "tokens/total": 2723984, "tokens/train_per_sec_per_gpu": 30.81, "tokens/trainable": 40997} +{"epoch": 0.35546875, "grad_norm": 0.15726213157176971, "learning_rate": 9.610167032937036e-05, "loss": 0.0043283673003315926, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00434, "step": 91, "tokens/total": 2754240, "tokens/train_per_sec_per_gpu": 36.39, "tokens/trainable": 41462} +{"epoch": 0.359375, "grad_norm": 1.0282626152038574, "learning_rate": 9.598262995928611e-05, "loss": 0.014904867857694626, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01502, "step": 92, "tokens/total": 2784672, "tokens/train_per_sec_per_gpu": 36.84, "tokens/trainable": 41927} +{"epoch": 0.36328125, "grad_norm": 0.2766452133655548, "learning_rate": 9.586188413468492e-05, "loss": 0.0071023497730493546, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00713, "step": 93, "tokens/total": 2815120, "tokens/train_per_sec_per_gpu": 34.85, "tokens/trainable": 42397} +{"epoch": 0.3671875, "grad_norm": 0.9832137823104858, "learning_rate": 9.57394378802934e-05, "loss": 0.011417705565690994, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.01148, "step": 94, "tokens/total": 2845600, "tokens/train_per_sec_per_gpu": 36.54, "tokens/trainable": 42883} +{"epoch": 0.37109375, "grad_norm": 0.4514453709125519, "learning_rate": 9.56152962916e-05, "loss": 0.007363555021584034, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00739, "step": 95, "tokens/total": 2875952, "tokens/train_per_sec_per_gpu": 38.7, "tokens/trainable": 43381} +{"epoch": 0.375, "grad_norm": 0.555672287940979, "learning_rate": 9.548946453464296e-05, "loss": 0.00891757383942604, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00896, "step": 96, "tokens/total": 2906288, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 43824} +{"epoch": 0.37890625, "grad_norm": 0.37105968594551086, "learning_rate": 9.53619478457953e-05, "loss": 0.007098545785993338, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00712, "step": 97, "tokens/total": 2936752, "tokens/train_per_sec_per_gpu": 31.48, "tokens/trainable": 44255} +{"epoch": 0.3828125, "grad_norm": 1.0755594968795776, "learning_rate": 9.523275153154695e-05, "loss": 0.028615131974220276, "memory/device_reserved (GiB)": 35.19, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.02903, "step": 98, "tokens/total": 2967072, "tokens/train_per_sec_per_gpu": 37.66, "tokens/trainable": 44739} +{"epoch": 0.38671875, "grad_norm": 0.9296021461486816, "learning_rate": 9.51018809682839e-05, "loss": 0.01639867201447487, "memory/device_reserved (GiB)": 35.76, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.01653, "step": 99, "tokens/total": 2997664, "tokens/train_per_sec_per_gpu": 36.17, "tokens/trainable": 45202} +{"epoch": 0.390625, "grad_norm": 0.6913011074066162, "learning_rate": 9.49693416020645e-05, "loss": 0.015926485881209373, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01605, "step": 100, "tokens/total": 3028032, "tokens/train_per_sec_per_gpu": 33.06, "tokens/trainable": 45646} +{"epoch": 0.39453125, "grad_norm": 0.3258785605430603, "learning_rate": 9.483513894839276e-05, "loss": 0.008693980053067207, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00873, "step": 101, "tokens/total": 3058272, "tokens/train_per_sec_per_gpu": 33.06, "tokens/trainable": 46089} +{"epoch": 0.3984375, "grad_norm": 1.0787935256958008, "learning_rate": 9.469927859198888e-05, "loss": 0.01954980194568634, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01974, "step": 102, "tokens/total": 3088848, "tokens/train_per_sec_per_gpu": 32.59, "tokens/trainable": 46547} +{"epoch": 0.40234375, "grad_norm": 0.9061365127563477, "learning_rate": 9.456176618655689e-05, "loss": 0.03529620170593262, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.03593, "step": 103, "tokens/total": 3119424, "tokens/train_per_sec_per_gpu": 31.97, "tokens/trainable": 47014} +{"epoch": 0.40625, "grad_norm": 2.671016216278076, "learning_rate": 9.442260745454927e-05, "loss": 0.02147374488413334, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.02171, "step": 104, "tokens/total": 3149696, "tokens/train_per_sec_per_gpu": 37.68, "tokens/trainable": 47503} +{"epoch": 0.41015625, "grad_norm": 0.26923975348472595, "learning_rate": 9.428180818692884e-05, "loss": 0.007999668829143047, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00803, "step": 105, "tokens/total": 3180064, "tokens/train_per_sec_per_gpu": 36.28, "tokens/trainable": 47996} +{"epoch": 0.4140625, "grad_norm": 0.7145718932151794, "learning_rate": 9.413937424292791e-05, "loss": 0.030988795682787895, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.03147, "step": 106, "tokens/total": 3208320, "tokens/train_per_sec_per_gpu": 34.04, "tokens/trainable": 48447} +{"epoch": 0.41796875, "grad_norm": 0.20594321191310883, "learning_rate": 9.399531154980424e-05, "loss": 0.007135331630706787, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00716, "step": 107, "tokens/total": 3238416, "tokens/train_per_sec_per_gpu": 30.69, "tokens/trainable": 48889} +{"epoch": 0.421875, "grad_norm": 0.2812410891056061, "learning_rate": 9.384962610259455e-05, "loss": 0.010571298189461231, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01063, "step": 108, "tokens/total": 3268784, "tokens/train_per_sec_per_gpu": 37.9, "tokens/trainable": 49360} +{"epoch": 0.42578125, "grad_norm": 0.3485014736652374, "learning_rate": 9.370232396386494e-05, "loss": 0.016001982614398003, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01613, "step": 109, "tokens/total": 3298912, "tokens/train_per_sec_per_gpu": 36.73, "tokens/trainable": 49827} +{"epoch": 0.4296875, "grad_norm": 1.474901795387268, "learning_rate": 9.355341126345868e-05, "loss": 0.024302365258336067, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.0246, "step": 110, "tokens/total": 3329232, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 50267} +{"epoch": 0.43359375, "grad_norm": 0.5579241514205933, "learning_rate": 9.340289419824107e-05, "loss": 0.017602810636162758, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01776, "step": 111, "tokens/total": 3359440, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 50711} +{"epoch": 0.4375, "grad_norm": 0.5616764426231384, "learning_rate": 9.325077903184159e-05, "loss": 0.015218488872051239, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.01533, "step": 112, "tokens/total": 3389472, "tokens/train_per_sec_per_gpu": 31.58, "tokens/trainable": 51157} +{"epoch": 0.44140625, "grad_norm": 0.38610321283340454, "learning_rate": 9.30970720943932e-05, "loss": 0.008289901539683342, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00832, "step": 113, "tokens/total": 3419920, "tokens/train_per_sec_per_gpu": 32.67, "tokens/trainable": 51613} +{"epoch": 0.4453125, "grad_norm": 0.3681856095790863, "learning_rate": 9.2941779782269e-05, "loss": 0.022977739572525024, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.02324, "step": 114, "tokens/total": 3450176, "tokens/train_per_sec_per_gpu": 29.07, "tokens/trainable": 52017} +{"epoch": 0.44921875, "grad_norm": 0.2453298270702362, "learning_rate": 9.278490855781596e-05, "loss": 0.004594211932271719, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0046, "step": 115, "tokens/total": 3480544, "tokens/train_per_sec_per_gpu": 36.22, "tokens/trainable": 52506} +{"epoch": 0.453125, "grad_norm": 0.37761810421943665, "learning_rate": 9.262646494908604e-05, "loss": 0.011891474947333336, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01196, "step": 116, "tokens/total": 3510848, "tokens/train_per_sec_per_gpu": 38.57, "tokens/trainable": 53008} +{"epoch": 0.45703125, "grad_norm": 1.953158974647522, "learning_rate": 9.246645554956457e-05, "loss": 0.006801537703722715, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00682, "step": 117, "tokens/total": 3541296, "tokens/train_per_sec_per_gpu": 35.38, "tokens/trainable": 53467} +{"epoch": 0.4609375, "grad_norm": 0.45078471302986145, "learning_rate": 9.230488701789578e-05, "loss": 0.017124727368354797, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.01727, "step": 118, "tokens/total": 3571936, "tokens/train_per_sec_per_gpu": 32.83, "tokens/trainable": 53923} +{"epoch": 0.46484375, "grad_norm": 0.6464744806289673, "learning_rate": 9.214176607760577e-05, "loss": 0.012542763724923134, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01262, "step": 119, "tokens/total": 3602000, "tokens/train_per_sec_per_gpu": 35.6, "tokens/trainable": 54363} +{"epoch": 0.46875, "grad_norm": 0.3361028730869293, "learning_rate": 9.197709951682268e-05, "loss": 0.004235772415995598, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00424, "step": 120, "tokens/total": 3632288, "tokens/train_per_sec_per_gpu": 35.26, "tokens/trainable": 54839} +{"epoch": 0.47265625, "grad_norm": 0.4455215036869049, "learning_rate": 9.181089418799428e-05, "loss": 0.004468773491680622, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00448, "step": 121, "tokens/total": 3662608, "tokens/train_per_sec_per_gpu": 32.62, "tokens/trainable": 55267} +{"epoch": 0.4765625, "grad_norm": 0.3547821044921875, "learning_rate": 9.164315700760271e-05, "loss": 0.010471776127815247, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.01053, "step": 122, "tokens/total": 3692864, "tokens/train_per_sec_per_gpu": 30.54, "tokens/trainable": 55700} +{"epoch": 0.48046875, "grad_norm": 0.7816572189331055, "learning_rate": 9.147389495587671e-05, "loss": 0.01556549035012722, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.01569, "step": 123, "tokens/total": 3722864, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 56144} +{"epoch": 0.484375, "grad_norm": 0.3188917338848114, "learning_rate": 9.130311507650116e-05, "loss": 0.0064091756939888, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00643, "step": 124, "tokens/total": 3753056, "tokens/train_per_sec_per_gpu": 33.13, "tokens/trainable": 56589} +{"epoch": 0.48828125, "grad_norm": 0.7098766565322876, "learning_rate": 9.113082447632394e-05, "loss": 0.026026558130979538, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02637, "step": 125, "tokens/total": 3783312, "tokens/train_per_sec_per_gpu": 36.6, "tokens/trainable": 57081} +{"epoch": 0.4921875, "grad_norm": 0.4195394217967987, "learning_rate": 9.09570303250602e-05, "loss": 0.01531171053647995, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01543, "step": 126, "tokens/total": 3813712, "tokens/train_per_sec_per_gpu": 32.83, "tokens/trainable": 57527} +{"epoch": 0.49609375, "grad_norm": 0.4466935992240906, "learning_rate": 9.078173985499394e-05, "loss": 0.014355337247252464, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01446, "step": 127, "tokens/total": 3844336, "tokens/train_per_sec_per_gpu": 34.84, "tokens/trainable": 58005} +{"epoch": 0.5, "grad_norm": 0.19709816575050354, "learning_rate": 9.060496036067713e-05, "loss": 0.006424291059374809, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.00644, "step": 128, "tokens/total": 3872688, "tokens/train_per_sec_per_gpu": 35.24, "tokens/trainable": 58454} +{"epoch": 0.50390625, "grad_norm": 0.18546900153160095, "learning_rate": 9.042669919862615e-05, "loss": 0.003899412928149104, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.64, "memory/max_allocated (GiB)": 33.64, "ppl": 1.00391, "step": 129, "tokens/total": 3902640, "tokens/train_per_sec_per_gpu": 29.88, "tokens/trainable": 58896} +{"epoch": 0.5078125, "grad_norm": 0.2095804512500763, "learning_rate": 9.024696378701557e-05, "loss": 0.005734554957598448, "memory/device_reserved (GiB)": 34.68, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00575, "step": 130, "tokens/total": 3933104, "tokens/train_per_sec_per_gpu": 35.88, "tokens/trainable": 59356} +{"epoch": 0.51171875, "grad_norm": 0.3548737168312073, "learning_rate": 9.006576160536948e-05, "loss": 0.010866682976484299, "memory/device_reserved (GiB)": 34.92, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.01093, "step": 131, "tokens/total": 3963184, "tokens/train_per_sec_per_gpu": 32.18, "tokens/trainable": 59792} +{"epoch": 0.515625, "grad_norm": 0.24394433200359344, "learning_rate": 8.988310019425035e-05, "loss": 0.004735369700938463, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00475, "step": 132, "tokens/total": 3993648, "tokens/train_per_sec_per_gpu": 35.71, "tokens/trainable": 60266} +{"epoch": 0.51953125, "grad_norm": 0.3678644597530365, "learning_rate": 8.969898715494506e-05, "loss": 0.007982447743415833, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00801, "step": 133, "tokens/total": 4024000, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 60759} +{"epoch": 0.5234375, "grad_norm": 0.6605305075645447, "learning_rate": 8.951343014914869e-05, "loss": 0.02931230701506138, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.02975, "step": 134, "tokens/total": 4052480, "tokens/train_per_sec_per_gpu": 31.78, "tokens/trainable": 61204} +{"epoch": 0.52734375, "grad_norm": 0.1193099245429039, "learning_rate": 8.932643689864568e-05, "loss": 0.0025630262680351734, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00257, "step": 135, "tokens/total": 4082864, "tokens/train_per_sec_per_gpu": 35.97, "tokens/trainable": 61684} +{"epoch": 0.53125, "grad_norm": 0.12960873544216156, "learning_rate": 8.913801518498845e-05, "loss": 0.0019581567030400038, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00196, "step": 136, "tokens/total": 4113040, "tokens/train_per_sec_per_gpu": 38.25, "tokens/trainable": 62159} +{"epoch": 0.53515625, "grad_norm": 0.34460127353668213, "learning_rate": 8.894817284917364e-05, "loss": 0.01051283534616232, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01057, "step": 137, "tokens/total": 4143344, "tokens/train_per_sec_per_gpu": 35.31, "tokens/trainable": 62649} +{"epoch": 0.5390625, "grad_norm": 0.20421279966831207, "learning_rate": 8.875691779131569e-05, "loss": 0.0035965177230536938, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.0036, "step": 138, "tokens/total": 4173648, "tokens/train_per_sec_per_gpu": 36.67, "tokens/trainable": 63100} +{"epoch": 0.54296875, "grad_norm": 0.30597925186157227, "learning_rate": 8.856425797031829e-05, "loss": 0.008190772496163845, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00822, "step": 139, "tokens/total": 4204080, "tokens/train_per_sec_per_gpu": 37.61, "tokens/trainable": 63588} +{"epoch": 0.546875, "grad_norm": 0.5688289999961853, "learning_rate": 8.837020140354295e-05, "loss": 0.022904174402356148, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.42, "memory/max_allocated (GiB)": 33.42, "ppl": 1.02317, "step": 140, "tokens/total": 4232400, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 64005} +{"epoch": 0.55078125, "grad_norm": 0.5879082083702087, "learning_rate": 8.817475616647554e-05, "loss": 0.013611489906907082, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.0137, "step": 141, "tokens/total": 4262624, "tokens/train_per_sec_per_gpu": 35.14, "tokens/trainable": 64462} +{"epoch": 0.5546875, "grad_norm": 0.31034204363822937, "learning_rate": 8.797793039239017e-05, "loss": 0.008315861225128174, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00835, "step": 142, "tokens/total": 4293168, "tokens/train_per_sec_per_gpu": 33.95, "tokens/trainable": 64900} +{"epoch": 0.55859375, "grad_norm": 0.7010869979858398, "learning_rate": 8.777973227201069e-05, "loss": 0.022431056946516037, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.02268, "step": 143, "tokens/total": 4323248, "tokens/train_per_sec_per_gpu": 37.57, "tokens/trainable": 65379} +{"epoch": 0.5625, "grad_norm": 0.2407914400100708, "learning_rate": 8.758017005316988e-05, "loss": 0.006207741796970367, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00623, "step": 144, "tokens/total": 4353344, "tokens/train_per_sec_per_gpu": 36.72, "tokens/trainable": 65845} +{"epoch": 0.56640625, "grad_norm": 0.21913351118564606, "learning_rate": 8.737925204046629e-05, "loss": 0.009031139314174652, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00907, "step": 145, "tokens/total": 4383584, "tokens/train_per_sec_per_gpu": 32.92, "tokens/trainable": 66285} +{"epoch": 0.5703125, "grad_norm": 0.39635229110717773, "learning_rate": 8.717698659491851e-05, "loss": 0.016251137480139732, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.01638, "step": 146, "tokens/total": 4413856, "tokens/train_per_sec_per_gpu": 31.48, "tokens/trainable": 66736} +{"epoch": 0.57421875, "grad_norm": 0.5958674550056458, "learning_rate": 8.697338213361735e-05, "loss": 0.018637506291270256, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01881, "step": 147, "tokens/total": 4444112, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 67150} +{"epoch": 0.578125, "grad_norm": 0.19332216680049896, "learning_rate": 8.676844712937552e-05, "loss": 0.006893862038850784, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00692, "step": 148, "tokens/total": 4474464, "tokens/train_per_sec_per_gpu": 35.09, "tokens/trainable": 67633} +{"epoch": 0.58203125, "grad_norm": 0.36499589681625366, "learning_rate": 8.656219011037509e-05, "loss": 0.010607698000967503, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01066, "step": 149, "tokens/total": 4504880, "tokens/train_per_sec_per_gpu": 33.87, "tokens/trainable": 68077} +{"epoch": 0.5859375, "grad_norm": 0.10280518978834152, "learning_rate": 8.63546196598125e-05, "loss": 0.002519096015021205, "memory/device_reserved (GiB)": 35.79, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00252, "step": 150, "tokens/total": 4534976, "tokens/train_per_sec_per_gpu": 33.26, "tokens/trainable": 68545} +{"epoch": 0.58984375, "grad_norm": 0.24021312594413757, "learning_rate": 8.614574441554145e-05, "loss": 0.006152496673166752, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00617, "step": 151, "tokens/total": 4565088, "tokens/train_per_sec_per_gpu": 30.26, "tokens/trainable": 69011} +{"epoch": 0.59375, "grad_norm": 0.5905681252479553, "learning_rate": 8.593557306971349e-05, "loss": 0.0134108392521739, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0135, "step": 152, "tokens/total": 4595600, "tokens/train_per_sec_per_gpu": 35.41, "tokens/trainable": 69465} +{"epoch": 0.59765625, "grad_norm": 0.5932232141494751, "learning_rate": 8.572411436841618e-05, "loss": 0.03526075929403305, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.03589, "step": 153, "tokens/total": 4626032, "tokens/train_per_sec_per_gpu": 33.9, "tokens/trainable": 69928} +{"epoch": 0.6015625, "grad_norm": 0.21574290096759796, "learning_rate": 8.551137711130922e-05, "loss": 0.0023245825432240963, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00233, "step": 154, "tokens/total": 4656352, "tokens/train_per_sec_per_gpu": 36.68, "tokens/trainable": 70395} +{"epoch": 0.60546875, "grad_norm": 0.08898661285638809, "learning_rate": 8.529737015125824e-05, "loss": 0.0007986892596818507, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0008, "step": 155, "tokens/total": 4684672, "tokens/train_per_sec_per_gpu": 35.61, "tokens/trainable": 70829} +{"epoch": 0.609375, "grad_norm": 0.1178978830575943, "learning_rate": 8.508210239396639e-05, "loss": 0.0014696650905534625, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00147, "step": 156, "tokens/total": 4715216, "tokens/train_per_sec_per_gpu": 33.9, "tokens/trainable": 71287} +{"epoch": 0.61328125, "grad_norm": 0.06257754564285278, "learning_rate": 8.486558279760375e-05, "loss": 0.000776763423345983, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00078, "step": 157, "tokens/total": 4745376, "tokens/train_per_sec_per_gpu": 35.05, "tokens/trainable": 71749} +{"epoch": 0.6171875, "grad_norm": 0.8163263201713562, "learning_rate": 8.464782037243449e-05, "loss": 0.033990029245615005, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.03457, "step": 158, "tokens/total": 4775776, "tokens/train_per_sec_per_gpu": 32.19, "tokens/trainable": 72185} +{"epoch": 0.62109375, "grad_norm": 0.014705101028084755, "learning_rate": 8.442882418044202e-05, "loss": 0.00028892228147014976, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00029, "step": 159, "tokens/total": 4805856, "tokens/train_per_sec_per_gpu": 30.53, "tokens/trainable": 72604} +{"epoch": 0.625, "grad_norm": 0.4126265048980713, "learning_rate": 8.420860333495179e-05, "loss": 0.019177839159965515, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01936, "step": 160, "tokens/total": 4836304, "tokens/train_per_sec_per_gpu": 31.58, "tokens/trainable": 73061} +{"epoch": 0.62890625, "grad_norm": 0.6193662285804749, "learning_rate": 8.398716700025208e-05, "loss": 0.008987972512841225, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00903, "step": 161, "tokens/total": 4866544, "tokens/train_per_sec_per_gpu": 29.97, "tokens/trainable": 73474} +{"epoch": 0.6328125, "grad_norm": 0.6351296901702881, "learning_rate": 8.376452439121266e-05, "loss": 0.01764395274221897, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0178, "step": 162, "tokens/total": 4897120, "tokens/train_per_sec_per_gpu": 32.68, "tokens/trainable": 73923} +{"epoch": 0.63671875, "grad_norm": 0.19809548556804657, "learning_rate": 8.354068477290124e-05, "loss": 0.002400235738605261, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0024, "step": 163, "tokens/total": 4927440, "tokens/train_per_sec_per_gpu": 35.37, "tokens/trainable": 74379} +{"epoch": 0.640625, "grad_norm": 0.17828311026096344, "learning_rate": 8.331565746019807e-05, "loss": 0.0019721905700862408, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00197, "step": 164, "tokens/total": 4958032, "tokens/train_per_sec_per_gpu": 33.46, "tokens/trainable": 74830} +{"epoch": 0.64453125, "grad_norm": 0.14491626620292664, "learning_rate": 8.308945181740812e-05, "loss": 0.0023866910487413406, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00239, "step": 165, "tokens/total": 4988368, "tokens/train_per_sec_per_gpu": 31.73, "tokens/trainable": 75262} +{"epoch": 0.6484375, "grad_norm": 0.11567407846450806, "learning_rate": 8.286207725787153e-05, "loss": 0.0012497061397880316, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00125, "step": 166, "tokens/total": 5018672, "tokens/train_per_sec_per_gpu": 31.84, "tokens/trainable": 75703} +{"epoch": 0.65234375, "grad_norm": 0.3614877164363861, "learning_rate": 8.263354324357182e-05, "loss": 0.010097509250044823, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.01015, "step": 167, "tokens/total": 5049344, "tokens/train_per_sec_per_gpu": 31.28, "tokens/trainable": 76164} +{"epoch": 0.65625, "grad_norm": 0.21049293875694275, "learning_rate": 8.240385928474219e-05, "loss": 0.005174943245947361, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00519, "step": 168, "tokens/total": 5079856, "tokens/train_per_sec_per_gpu": 34.46, "tokens/trainable": 76623} +{"epoch": 0.66015625, "grad_norm": 0.16444046795368195, "learning_rate": 8.217303493946967e-05, "loss": 0.0028489381074905396, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00285, "step": 169, "tokens/total": 5110192, "tokens/train_per_sec_per_gpu": 34.39, "tokens/trainable": 77047} +{"epoch": 0.6640625, "grad_norm": 0.29660263657569885, "learning_rate": 8.194107981329746e-05, "loss": 0.020341932773590088, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02055, "step": 170, "tokens/total": 5140592, "tokens/train_per_sec_per_gpu": 34.08, "tokens/trainable": 77496} +{"epoch": 0.66796875, "grad_norm": 0.2912541627883911, "learning_rate": 8.170800355882518e-05, "loss": 0.013131296262145042, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01322, "step": 171, "tokens/total": 5170848, "tokens/train_per_sec_per_gpu": 33.24, "tokens/trainable": 77956} +{"epoch": 0.671875, "grad_norm": 0.14199896156787872, "learning_rate": 8.147381587530713e-05, "loss": 0.002253658836707473, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00226, "step": 172, "tokens/total": 5201280, "tokens/train_per_sec_per_gpu": 29.74, "tokens/trainable": 78410} +{"epoch": 0.67578125, "grad_norm": 0.6508966684341431, "learning_rate": 8.123852650824877e-05, "loss": 0.01731065846979618, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01746, "step": 173, "tokens/total": 5231712, "tokens/train_per_sec_per_gpu": 32.99, "tokens/trainable": 78849} +{"epoch": 0.6796875, "grad_norm": 0.22890180349349976, "learning_rate": 8.100214524900103e-05, "loss": 0.013438408263027668, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.62, "memory/max_allocated (GiB)": 33.62, "ppl": 1.01353, "step": 174, "tokens/total": 5261440, "tokens/train_per_sec_per_gpu": 30.1, "tokens/trainable": 79266} +{"epoch": 0.68359375, "grad_norm": 0.034085191786289215, "learning_rate": 8.076468193435301e-05, "loss": 0.0010016995947808027, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.001, "step": 175, "tokens/total": 5291728, "tokens/train_per_sec_per_gpu": 32.5, "tokens/trainable": 79736} +{"epoch": 0.6875, "grad_norm": 0.1346512883901596, "learning_rate": 8.052614644612253e-05, "loss": 0.0026639762800186872, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00267, "step": 176, "tokens/total": 5322240, "tokens/train_per_sec_per_gpu": 36.14, "tokens/trainable": 80214} +{"epoch": 0.69140625, "grad_norm": 0.2848314940929413, "learning_rate": 8.028654871074489e-05, "loss": 0.008686527609825134, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00872, "step": 177, "tokens/total": 5352576, "tokens/train_per_sec_per_gpu": 32.48, "tokens/trainable": 80661} +{"epoch": 0.6953125, "grad_norm": 0.037947915494441986, "learning_rate": 8.004589869885986e-05, "loss": 0.0010221066186204553, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00102, "step": 178, "tokens/total": 5382784, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 81129} +{"epoch": 0.69921875, "grad_norm": 0.1301245242357254, "learning_rate": 7.980420642489674e-05, "loss": 0.0017726586665958166, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00177, "step": 179, "tokens/total": 5412784, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 81576} +{"epoch": 0.703125, "grad_norm": 0.16076700389385223, "learning_rate": 7.95614819466576e-05, "loss": 0.0021704614628106356, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00217, "step": 180, "tokens/total": 5442960, "tokens/train_per_sec_per_gpu": 34.55, "tokens/trainable": 82049} +{"epoch": 0.70703125, "grad_norm": 0.5916904807090759, "learning_rate": 7.931773536489872e-05, "loss": 0.018786190077662468, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01896, "step": 181, "tokens/total": 5473216, "tokens/train_per_sec_per_gpu": 38.83, "tokens/trainable": 82566} +{"epoch": 0.7109375, "grad_norm": 0.140871062874794, "learning_rate": 7.907297682291035e-05, "loss": 0.002907736459746957, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00291, "step": 182, "tokens/total": 5503568, "tokens/train_per_sec_per_gpu": 33.93, "tokens/trainable": 83054} +{"epoch": 0.71484375, "grad_norm": 0.3186001181602478, "learning_rate": 7.882721650609442e-05, "loss": 0.002825568662956357, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00283, "step": 183, "tokens/total": 5533760, "tokens/train_per_sec_per_gpu": 32.79, "tokens/trainable": 83501} +{"epoch": 0.71875, "grad_norm": 0.10091359168291092, "learning_rate": 7.85804646415409e-05, "loss": 0.000788951525464654, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00079, "step": 184, "tokens/total": 5564240, "tokens/train_per_sec_per_gpu": 35.52, "tokens/trainable": 84007} +{"epoch": 0.72265625, "grad_norm": 0.29195457696914673, "learning_rate": 7.833273149760207e-05, "loss": 0.006313872057944536, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00633, "step": 185, "tokens/total": 5594448, "tokens/train_per_sec_per_gpu": 35.57, "tokens/trainable": 84491} +{"epoch": 0.7265625, "grad_norm": 0.5876129269599915, "learning_rate": 7.808402738346527e-05, "loss": 0.01837407797574997, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01854, "step": 186, "tokens/total": 5624816, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 84952} +{"epoch": 0.73046875, "grad_norm": 0.4564031660556793, "learning_rate": 7.783436264872382e-05, "loss": 0.00939339492470026, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00944, "step": 187, "tokens/total": 5655104, "tokens/train_per_sec_per_gpu": 32.93, "tokens/trainable": 85400} +{"epoch": 0.734375, "grad_norm": 0.018894854933023453, "learning_rate": 7.758374768294647e-05, "loss": 0.0005394529434852302, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00054, "step": 188, "tokens/total": 5685360, "tokens/train_per_sec_per_gpu": 33.98, "tokens/trainable": 85889} +{"epoch": 0.73828125, "grad_norm": 0.11897671222686768, "learning_rate": 7.733219291524489e-05, "loss": 0.0019742052536457777, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00198, "step": 189, "tokens/total": 5715648, "tokens/train_per_sec_per_gpu": 36.25, "tokens/trainable": 86396} +{"epoch": 0.7421875, "grad_norm": 0.24015508592128754, "learning_rate": 7.707970881383977e-05, "loss": 0.007755734957754612, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00779, "step": 190, "tokens/total": 5746240, "tokens/train_per_sec_per_gpu": 33.86, "tokens/trainable": 86867} +{"epoch": 0.74609375, "grad_norm": 0.2647373378276825, "learning_rate": 7.682630588562518e-05, "loss": 0.017713403329253197, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.01787, "step": 191, "tokens/total": 5776816, "tokens/train_per_sec_per_gpu": 36.16, "tokens/trainable": 87314} +{"epoch": 0.75, "grad_norm": 0.2803303599357605, "learning_rate": 7.657199467573129e-05, "loss": 0.008597557432949543, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00863, "step": 192, "tokens/total": 5807248, "tokens/train_per_sec_per_gpu": 30.38, "tokens/trainable": 87736} +{"epoch": 0.75390625, "grad_norm": 0.045643068850040436, "learning_rate": 7.631678576708561e-05, "loss": 0.0008268685778602958, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00083, "step": 193, "tokens/total": 5837680, "tokens/train_per_sec_per_gpu": 35.05, "tokens/trainable": 88219} +{"epoch": 0.7578125, "grad_norm": 0.49878475069999695, "learning_rate": 7.606068977997255e-05, "loss": 0.008574062027037144, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00861, "step": 194, "tokens/total": 5868000, "tokens/train_per_sec_per_gpu": 36.12, "tokens/trainable": 88699} +{"epoch": 0.76171875, "grad_norm": 0.0975877121090889, "learning_rate": 7.580371737159148e-05, "loss": 0.0020896620117127895, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.38, "memory/max_allocated (GiB)": 33.38, "ppl": 1.00209, "step": 195, "tokens/total": 5896224, "tokens/train_per_sec_per_gpu": 30.38, "tokens/trainable": 89132} +{"epoch": 0.765625, "grad_norm": 0.18829824030399323, "learning_rate": 7.554587923561324e-05, "loss": 0.004770598374307156, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00478, "step": 196, "tokens/total": 5926624, "tokens/train_per_sec_per_gpu": 37.22, "tokens/trainable": 89594} +{"epoch": 0.76953125, "grad_norm": 0.11453713476657867, "learning_rate": 7.528718610173511e-05, "loss": 0.0024538300931453705, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.34, "memory/max_allocated (GiB)": 33.34, "ppl": 1.00246, "step": 197, "tokens/total": 5954944, "tokens/train_per_sec_per_gpu": 38.11, "tokens/trainable": 90042} +{"epoch": 0.7734375, "grad_norm": 0.3371991217136383, "learning_rate": 7.502764873523431e-05, "loss": 0.012420275248587132, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0125, "step": 198, "tokens/total": 5985200, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 90503} +{"epoch": 0.77734375, "grad_norm": 0.044131435453891754, "learning_rate": 7.476727793652011e-05, "loss": 0.0006442728918045759, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 34.0, "memory/max_allocated (GiB)": 34.0, "ppl": 1.00064, "step": 199, "tokens/total": 6015952, "tokens/train_per_sec_per_gpu": 36.52, "tokens/trainable": 90982} +{"epoch": 0.78125, "grad_norm": 0.11669042706489563, "learning_rate": 7.450608454068415e-05, "loss": 0.0016041958006098866, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00161, "step": 200, "tokens/total": 6046208, "tokens/train_per_sec_per_gpu": 39.32, "tokens/trainable": 91466} +{"epoch": 0.78515625, "grad_norm": 0.030662450939416885, "learning_rate": 7.424407941704987e-05, "loss": 0.0007836977019906044, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00078, "step": 201, "tokens/total": 6076208, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 91912} +{"epoch": 0.7890625, "grad_norm": 0.16404050588607788, "learning_rate": 7.398127346871986e-05, "loss": 0.0035354122519493103, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00354, "step": 202, "tokens/total": 6106560, "tokens/train_per_sec_per_gpu": 37.84, "tokens/trainable": 92385} +{"epoch": 0.79296875, "grad_norm": 0.3355195224285126, "learning_rate": 7.371767763212238e-05, "loss": 0.029696868732571602, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.03014, "step": 203, "tokens/total": 6136640, "tokens/train_per_sec_per_gpu": 36.55, "tokens/trainable": 92842} +{"epoch": 0.796875, "grad_norm": 0.2913725972175598, "learning_rate": 7.345330287655617e-05, "loss": 0.004002463538199663, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00401, "step": 204, "tokens/total": 6167184, "tokens/train_per_sec_per_gpu": 34.02, "tokens/trainable": 93323} +{"epoch": 0.80078125, "grad_norm": 0.43749040365219116, "learning_rate": 7.31881602037339e-05, "loss": 0.013340720906853676, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.01343, "step": 205, "tokens/total": 6197376, "tokens/train_per_sec_per_gpu": 34.8, "tokens/trainable": 93756} +{"epoch": 0.8046875, "grad_norm": 0.4968360960483551, "learning_rate": 7.29222606473245e-05, "loss": 0.01012748945504427, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01018, "step": 206, "tokens/total": 6227600, "tokens/train_per_sec_per_gpu": 37.58, "tokens/trainable": 94235} +{"epoch": 0.80859375, "grad_norm": 0.06706669926643372, "learning_rate": 7.265561527249383e-05, "loss": 0.0005324441008269787, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00053, "step": 207, "tokens/total": 6257760, "tokens/train_per_sec_per_gpu": 36.54, "tokens/trainable": 94733} +{"epoch": 0.8125, "grad_norm": 0.14510789513587952, "learning_rate": 7.238823517544436e-05, "loss": 0.0053801448084414005, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00539, "step": 208, "tokens/total": 6288048, "tokens/train_per_sec_per_gpu": 33.74, "tokens/trainable": 95218} +{"epoch": 0.81640625, "grad_norm": 0.07251987606287003, "learning_rate": 7.212013148295333e-05, "loss": 0.0016005284851416945, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.0016, "step": 209, "tokens/total": 6318688, "tokens/train_per_sec_per_gpu": 29.23, "tokens/trainable": 95619} +{"epoch": 0.8203125, "grad_norm": 0.026614652946591377, "learning_rate": 7.185131535190975e-05, "loss": 0.0005790021969005466, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00058, "step": 210, "tokens/total": 6348944, "tokens/train_per_sec_per_gpu": 35.4, "tokens/trainable": 96076} +{"epoch": 0.82421875, "grad_norm": 0.04637908935546875, "learning_rate": 7.158179796885005e-05, "loss": 0.0009742251713760197, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00097, "step": 211, "tokens/total": 6378960, "tokens/train_per_sec_per_gpu": 35.36, "tokens/trainable": 96505} +{"epoch": 0.828125, "grad_norm": 0.07815160602331161, "learning_rate": 7.131159054949273e-05, "loss": 0.0010261747520416975, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00103, "step": 212, "tokens/total": 6409472, "tokens/train_per_sec_per_gpu": 32.67, "tokens/trainable": 96930} +{"epoch": 0.83203125, "grad_norm": 0.032491739839315414, "learning_rate": 7.104070433827139e-05, "loss": 0.0007142109097912908, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00071, "step": 213, "tokens/total": 6439904, "tokens/train_per_sec_per_gpu": 32.72, "tokens/trainable": 97361} +{"epoch": 0.8359375, "grad_norm": 0.26459500193595886, "learning_rate": 7.076915060786705e-05, "loss": 0.00510193407535553, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00511, "step": 214, "tokens/total": 6470176, "tokens/train_per_sec_per_gpu": 36.36, "tokens/trainable": 97858} +{"epoch": 0.83984375, "grad_norm": 0.4014389216899872, "learning_rate": 7.049694065873882e-05, "loss": 0.01917858049273491, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01936, "step": 215, "tokens/total": 6500336, "tokens/train_per_sec_per_gpu": 37.49, "tokens/trainable": 98308} +{"epoch": 0.84375, "grad_norm": 0.017172295600175858, "learning_rate": 7.022408581865382e-05, "loss": 0.00041207089088857174, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00041, "step": 216, "tokens/total": 6530688, "tokens/train_per_sec_per_gpu": 36.38, "tokens/trainable": 98781} +{"epoch": 0.84765625, "grad_norm": 0.21681398153305054, "learning_rate": 6.99505974422157e-05, "loss": 0.0024940790608525276, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.0025, "step": 217, "tokens/total": 6561040, "tokens/train_per_sec_per_gpu": 36.75, "tokens/trainable": 99266} +{"epoch": 0.8515625, "grad_norm": 0.15546201169490814, "learning_rate": 6.967648691039213e-05, "loss": 0.0021202065981924534, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00212, "step": 218, "tokens/total": 6589392, "tokens/train_per_sec_per_gpu": 39.67, "tokens/trainable": 99731} +{"epoch": 0.85546875, "grad_norm": 0.14469008147716522, "learning_rate": 6.940176563004123e-05, "loss": 0.001545824226923287, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00155, "step": 219, "tokens/total": 6619824, "tokens/train_per_sec_per_gpu": 31.74, "tokens/trainable": 100164} +{"epoch": 0.859375, "grad_norm": 0.36150020360946655, "learning_rate": 6.912644503343682e-05, "loss": 0.015608995221555233, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01573, "step": 220, "tokens/total": 6650224, "tokens/train_per_sec_per_gpu": 36.18, "tokens/trainable": 100620} +{"epoch": 0.86328125, "grad_norm": 0.05116768181324005, "learning_rate": 6.885053657779273e-05, "loss": 0.0009821112034842372, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00098, "step": 221, "tokens/total": 6680384, "tokens/train_per_sec_per_gpu": 32.94, "tokens/trainable": 101080} +{"epoch": 0.8671875, "grad_norm": 1.1284459829330444, "learning_rate": 6.857405174478604e-05, "loss": 0.016592765226960182, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01673, "step": 222, "tokens/total": 6710896, "tokens/train_per_sec_per_gpu": 37.76, "tokens/trainable": 101546} +{"epoch": 0.87109375, "grad_norm": 0.0270449947565794, "learning_rate": 6.82970020400792e-05, "loss": 0.0005479598185047507, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00055, "step": 223, "tokens/total": 6741280, "tokens/train_per_sec_per_gpu": 31.55, "tokens/trainable": 101987} +{"epoch": 0.875, "grad_norm": 0.28190311789512634, "learning_rate": 6.801939899284132e-05, "loss": 0.016690485179424286, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01683, "step": 224, "tokens/total": 6771488, "tokens/train_per_sec_per_gpu": 35.47, "tokens/trainable": 102435} +{"epoch": 0.87890625, "grad_norm": 0.31064996123313904, "learning_rate": 6.774125415526827e-05, "loss": 0.007865586318075657, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0079, "step": 225, "tokens/total": 6801744, "tokens/train_per_sec_per_gpu": 29.1, "tokens/trainable": 102843} +{"epoch": 0.8828125, "grad_norm": 0.21806642413139343, "learning_rate": 6.746257910210214e-05, "loss": 0.006044697016477585, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00606, "step": 226, "tokens/total": 6832432, "tokens/train_per_sec_per_gpu": 37.25, "tokens/trainable": 103342} +{"epoch": 0.88671875, "grad_norm": 0.09062594920396805, "learning_rate": 6.718338543014937e-05, "loss": 0.001718292129226029, "memory/device_reserved (GiB)": 35.71, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00172, "step": 227, "tokens/total": 6862448, "tokens/train_per_sec_per_gpu": 40.76, "tokens/trainable": 103844} +{"epoch": 0.890625, "grad_norm": 0.9462765455245972, "learning_rate": 6.69036847577983e-05, "loss": 0.01898857392370701, "memory/device_reserved (GiB)": 35.77, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01917, "step": 228, "tokens/total": 6892832, "tokens/train_per_sec_per_gpu": 36.17, "tokens/trainable": 104307} +{"epoch": 0.89453125, "grad_norm": 0.0640437975525856, "learning_rate": 6.662348872453553e-05, "loss": 0.0018462492153048515, "memory/device_reserved (GiB)": 35.77, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00185, "step": 229, "tokens/total": 6922832, "tokens/train_per_sec_per_gpu": 33.47, "tokens/trainable": 104772} +{"epoch": 0.8984375, "grad_norm": 0.11650252342224121, "learning_rate": 6.63428089904618e-05, "loss": 0.004365470260381699, "memory/device_reserved (GiB)": 35.77, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00438, "step": 230, "tokens/total": 6953104, "tokens/train_per_sec_per_gpu": 36.8, "tokens/trainable": 105240} +{"epoch": 0.90234375, "grad_norm": 0.07681138068437576, "learning_rate": 6.60616572358065e-05, "loss": 0.002444313373416662, "memory/device_reserved (GiB)": 35.77, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00245, "step": 231, "tokens/total": 6983440, "tokens/train_per_sec_per_gpu": 33.74, "tokens/trainable": 105703} +{"epoch": 0.90625, "grad_norm": 0.20511816442012787, "learning_rate": 6.578004516044172e-05, "loss": 0.004132373258471489, "memory/device_reserved (GiB)": 35.77, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00414, "step": 232, "tokens/total": 7013840, "tokens/train_per_sec_per_gpu": 32.96, "tokens/trainable": 106179} +{"epoch": 0.91015625, "grad_norm": 0.11913148313760757, "learning_rate": 6.549798448339548e-05, "loss": 0.0025868406519293785, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00259, "step": 233, "tokens/total": 7044240, "tokens/train_per_sec_per_gpu": 37.97, "tokens/trainable": 106683} +{"epoch": 0.9140625, "grad_norm": 0.03754379227757454, "learning_rate": 6.521548694236384e-05, "loss": 0.0014621062437072396, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00146, "step": 234, "tokens/total": 7074224, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 107121} +{"epoch": 0.91796875, "grad_norm": 0.09151529520750046, "learning_rate": 6.493256429322259e-05, "loss": 0.002102814381942153, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00211, "step": 235, "tokens/total": 7104672, "tokens/train_per_sec_per_gpu": 32.69, "tokens/trainable": 107561} +{"epoch": 0.921875, "grad_norm": 0.09312661737203598, "learning_rate": 6.464922830953799e-05, "loss": 0.002275613136589527, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00228, "step": 236, "tokens/total": 7134784, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 108018} +{"epoch": 0.92578125, "grad_norm": 0.02780505083501339, "learning_rate": 6.436549078207688e-05, "loss": 0.0009309989400207996, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00093, "step": 237, "tokens/total": 7164960, "tokens/train_per_sec_per_gpu": 33.28, "tokens/trainable": 108446} +{"epoch": 0.9296875, "grad_norm": 0.014633426442742348, "learning_rate": 6.408136351831592e-05, "loss": 0.0004528576973825693, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00045, "step": 238, "tokens/total": 7195344, "tokens/train_per_sec_per_gpu": 29.06, "tokens/trainable": 108866} +{"epoch": 0.93359375, "grad_norm": 0.12943324446678162, "learning_rate": 6.379685834195036e-05, "loss": 0.001399983186274767, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0014, "step": 239, "tokens/total": 7225680, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 109316} +{"epoch": 0.9375, "grad_norm": 0.1406504064798355, "learning_rate": 6.351198709240186e-05, "loss": 0.002863554749637842, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00287, "step": 240, "tokens/total": 7256112, "tokens/train_per_sec_per_gpu": 33.28, "tokens/trainable": 109787} +{"epoch": 0.94140625, "grad_norm": 0.016691425815224648, "learning_rate": 6.32267616243259e-05, "loss": 0.0003821298014372587, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00038, "step": 241, "tokens/total": 7286384, "tokens/train_per_sec_per_gpu": 38.35, "tokens/trainable": 110246} +{"epoch": 0.9453125, "grad_norm": 0.038080595433712006, "learning_rate": 6.294119380711849e-05, "loss": 0.000662593636661768, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00066, "step": 242, "tokens/total": 7316688, "tokens/train_per_sec_per_gpu": 29.16, "tokens/trainable": 110667} +{"epoch": 0.94921875, "grad_norm": 0.2705026865005493, "learning_rate": 6.265529552442209e-05, "loss": 0.010765114799141884, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01082, "step": 243, "tokens/total": 7346992, "tokens/train_per_sec_per_gpu": 38.11, "tokens/trainable": 111154} +{"epoch": 0.953125, "grad_norm": 0.007892047055065632, "learning_rate": 6.236907867363127e-05, "loss": 0.00019494083244353533, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00019, "step": 244, "tokens/total": 7377632, "tokens/train_per_sec_per_gpu": 32.97, "tokens/trainable": 111611} +{"epoch": 0.95703125, "grad_norm": 0.03014756739139557, "learning_rate": 6.208255516539749e-05, "loss": 0.0004936520708724856, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00049, "step": 245, "tokens/total": 7407872, "tokens/train_per_sec_per_gpu": 34.19, "tokens/trainable": 112102} +{"epoch": 0.9609375, "grad_norm": 0.17086820304393768, "learning_rate": 6.179573692313344e-05, "loss": 0.0017580927815288305, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00176, "step": 246, "tokens/total": 7438448, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 112569} +{"epoch": 0.96484375, "grad_norm": 0.004760230891406536, "learning_rate": 6.150863588251694e-05, "loss": 0.00013790707453154027, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00014, "step": 247, "tokens/total": 7466768, "tokens/train_per_sec_per_gpu": 37.66, "tokens/trainable": 113023} +{"epoch": 0.96875, "grad_norm": 0.02571815438568592, "learning_rate": 6.122126399099419e-05, "loss": 0.0004854568687733263, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00049, "step": 248, "tokens/total": 7497360, "tokens/train_per_sec_per_gpu": 35.24, "tokens/trainable": 113464} +{"epoch": 0.97265625, "grad_norm": 5.738894462585449, "learning_rate": 6.0933633207282615e-05, "loss": 0.0020567248575389385, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00206, "step": 249, "tokens/total": 7527504, "tokens/train_per_sec_per_gpu": 32.69, "tokens/trainable": 113908} +{"epoch": 0.9765625, "grad_norm": 0.1298590898513794, "learning_rate": 6.064575550087316e-05, "loss": 0.001127784256823361, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00113, "step": 250, "tokens/total": 7555808, "tokens/train_per_sec_per_gpu": 35.31, "tokens/trainable": 114348} +{"epoch": 0.98046875, "grad_norm": 0.389615923166275, "learning_rate": 6.0357642851532245e-05, "loss": 0.0020728919189423323, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00208, "step": 251, "tokens/total": 7586320, "tokens/train_per_sec_per_gpu": 35.2, "tokens/trainable": 114830} +{"epoch": 0.984375, "grad_norm": 0.011385864578187466, "learning_rate": 6.0069307248803294e-05, "loss": 0.0001549161970615387, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00015, "step": 252, "tokens/total": 7616576, "tokens/train_per_sec_per_gpu": 35.18, "tokens/trainable": 115278} +{"epoch": 0.98828125, "grad_norm": 0.041026510298252106, "learning_rate": 5.9780760691507635e-05, "loss": 0.0002571218938101083, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00026, "step": 253, "tokens/total": 7647152, "tokens/train_per_sec_per_gpu": 32.26, "tokens/trainable": 115699} +{"epoch": 0.9921875, "grad_norm": 0.053424153476953506, "learning_rate": 5.9492015187245334e-05, "loss": 0.0005855828057974577, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00059, "step": 254, "tokens/total": 7677584, "tokens/train_per_sec_per_gpu": 30.78, "tokens/trainable": 116133} +{"epoch": 0.99609375, "grad_norm": 0.003189819399267435, "learning_rate": 5.920308275189541e-05, "loss": 6.202489021234214e-05, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00006, "step": 255, "tokens/total": 7708016, "tokens/train_per_sec_per_gpu": 36.81, "tokens/trainable": 116591} +{"epoch": 1.0, "grad_norm": 0.024073874577879906, "learning_rate": 5.8913975409115874e-05, "loss": 0.00023448803403880447, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00023, "step": 256, "tokens/total": 7738608, "tokens/train_per_sec_per_gpu": 32.08, "tokens/trainable": 117039} +{"epoch": 1.00390625, "grad_norm": 0.37222179770469666, "learning_rate": 5.8624705189843395e-05, "loss": 0.004106060601770878, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00411, "step": 257, "tokens/total": 7768944, "tokens/train_per_sec_per_gpu": 32.26, "tokens/trainable": 117495} +{"epoch": 1.0078125, "grad_norm": 0.6812037825584412, "learning_rate": 5.833528413179249e-05, "loss": 0.009335177019238472, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00938, "step": 258, "tokens/total": 7799232, "tokens/train_per_sec_per_gpu": 30.54, "tokens/trainable": 117925} +{"epoch": 1.01171875, "grad_norm": 0.003316260641440749, "learning_rate": 5.80457242789548e-05, "loss": 7.446527888532728e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00007, "step": 259, "tokens/total": 7829568, "tokens/train_per_sec_per_gpu": 31.65, "tokens/trainable": 118373} +{"epoch": 1.015625, "grad_norm": 0.016280459240078926, "learning_rate": 5.77560376810977e-05, "loss": 0.0001550098677398637, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00016, "step": 260, "tokens/total": 7860000, "tokens/train_per_sec_per_gpu": 31.53, "tokens/trainable": 118816} +{"epoch": 1.01953125, "grad_norm": 0.016491129994392395, "learning_rate": 5.7466236393263005e-05, "loss": 0.0001946197880897671, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00019, "step": 261, "tokens/total": 7890464, "tokens/train_per_sec_per_gpu": 33.83, "tokens/trainable": 119263} +{"epoch": 1.0234375, "grad_norm": 0.0688014104962349, "learning_rate": 5.717633247526522e-05, "loss": 0.0002748112892732024, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00027, "step": 262, "tokens/total": 7920944, "tokens/train_per_sec_per_gpu": 33.77, "tokens/trainable": 119772} +{"epoch": 1.02734375, "grad_norm": 0.3654443919658661, "learning_rate": 5.688633799118971e-05, "loss": 0.005442210007458925, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00546, "step": 263, "tokens/total": 7951296, "tokens/train_per_sec_per_gpu": 33.87, "tokens/trainable": 120269} +{"epoch": 1.03125, "grad_norm": 0.11719754338264465, "learning_rate": 5.659626500889066e-05, "loss": 0.0006240052171051502, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00062, "step": 264, "tokens/total": 7981856, "tokens/train_per_sec_per_gpu": 37.09, "tokens/trainable": 120746} +{"epoch": 1.03515625, "grad_norm": 0.042248908430337906, "learning_rate": 5.6306125599488905e-05, "loss": 0.000372910697478801, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00037, "step": 265, "tokens/total": 8012192, "tokens/train_per_sec_per_gpu": 38.31, "tokens/trainable": 121204} +{"epoch": 1.0390625, "grad_norm": 0.20307184755802155, "learning_rate": 5.601593183686955e-05, "loss": 0.0017405826365575194, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00174, "step": 266, "tokens/total": 8042736, "tokens/train_per_sec_per_gpu": 31.3, "tokens/trainable": 121670} +{"epoch": 1.04296875, "grad_norm": 0.0066534546203911304, "learning_rate": 5.572569579717961e-05, "loss": 0.00011396329500712454, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00011, "step": 267, "tokens/total": 8073024, "tokens/train_per_sec_per_gpu": 40.1, "tokens/trainable": 122183} +{"epoch": 1.046875, "grad_norm": 0.056601718068122864, "learning_rate": 5.543542955832538e-05, "loss": 0.0008997474797070026, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.0009, "step": 268, "tokens/total": 8103280, "tokens/train_per_sec_per_gpu": 35.43, "tokens/trainable": 122645} +{"epoch": 1.05078125, "grad_norm": 0.08372552692890167, "learning_rate": 5.514514519946986e-05, "loss": 0.0007820962928235531, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00078, "step": 269, "tokens/total": 8133440, "tokens/train_per_sec_per_gpu": 33.12, "tokens/trainable": 123105} +{"epoch": 1.0546875, "grad_norm": 0.2750205397605896, "learning_rate": 5.485485480053015e-05, "loss": 0.002154015703126788, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00216, "step": 270, "tokens/total": 8163920, "tokens/train_per_sec_per_gpu": 36.26, "tokens/trainable": 123595} +{"epoch": 1.05859375, "grad_norm": 0.6561075448989868, "learning_rate": 5.4564570441674645e-05, "loss": 0.006013512145727873, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00603, "step": 271, "tokens/total": 8194032, "tokens/train_per_sec_per_gpu": 37.69, "tokens/trainable": 124077} +{"epoch": 1.0625, "grad_norm": 0.1100449338555336, "learning_rate": 5.42743042028204e-05, "loss": 0.0009246943518519402, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00093, "step": 272, "tokens/total": 8224320, "tokens/train_per_sec_per_gpu": 34.74, "tokens/trainable": 124547} +{"epoch": 1.06640625, "grad_norm": 0.026166923344135284, "learning_rate": 5.3984068163130464e-05, "loss": 0.00030680614872835577, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00031, "step": 273, "tokens/total": 8254368, "tokens/train_per_sec_per_gpu": 34.04, "tokens/trainable": 124977} +{"epoch": 1.0703125, "grad_norm": 0.00845333095639944, "learning_rate": 5.369387440051111e-05, "loss": 0.00012944119225721806, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00013, "step": 274, "tokens/total": 8284944, "tokens/train_per_sec_per_gpu": 34.59, "tokens/trainable": 125438} +{"epoch": 1.07421875, "grad_norm": 0.5315051674842834, "learning_rate": 5.340373499110935e-05, "loss": 0.003738347440958023, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00375, "step": 275, "tokens/total": 8315584, "tokens/train_per_sec_per_gpu": 31.49, "tokens/trainable": 125889} +{"epoch": 1.078125, "grad_norm": 0.9425964951515198, "learning_rate": 5.3113662008810304e-05, "loss": 0.005198127124458551, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00521, "step": 276, "tokens/total": 8345952, "tokens/train_per_sec_per_gpu": 32.91, "tokens/trainable": 126363} +{"epoch": 1.08203125, "grad_norm": 0.0510481633245945, "learning_rate": 5.282366752473479e-05, "loss": 0.000684294558595866, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00068, "step": 277, "tokens/total": 8376416, "tokens/train_per_sec_per_gpu": 34.05, "tokens/trainable": 126828} +{"epoch": 1.0859375, "grad_norm": 0.12819825112819672, "learning_rate": 5.2533763606737005e-05, "loss": 0.0007626033038832247, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00076, "step": 278, "tokens/total": 8406640, "tokens/train_per_sec_per_gpu": 38.25, "tokens/trainable": 127331} +{"epoch": 1.08984375, "grad_norm": 0.1269499808549881, "learning_rate": 5.224396231890232e-05, "loss": 0.0010540171060711145, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00105, "step": 279, "tokens/total": 8436784, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 127798} +{"epoch": 1.09375, "grad_norm": 0.46639901399612427, "learning_rate": 5.195427572104522e-05, "loss": 0.01665615662932396, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0168, "step": 280, "tokens/total": 8467152, "tokens/train_per_sec_per_gpu": 32.87, "tokens/trainable": 128247} +{"epoch": 1.09765625, "grad_norm": 0.5340579748153687, "learning_rate": 5.166471586820751e-05, "loss": 0.016249053180217743, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01638, "step": 281, "tokens/total": 8497424, "tokens/train_per_sec_per_gpu": 36.24, "tokens/trainable": 128729} +{"epoch": 1.1015625, "grad_norm": 0.6115930676460266, "learning_rate": 5.1375294810156615e-05, "loss": 0.015617613680660725, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01574, "step": 282, "tokens/total": 8527616, "tokens/train_per_sec_per_gpu": 35.15, "tokens/trainable": 129203} +{"epoch": 1.10546875, "grad_norm": 0.4472820460796356, "learning_rate": 5.1086024590884144e-05, "loss": 0.005680762697011232, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0057, "step": 283, "tokens/total": 8557984, "tokens/train_per_sec_per_gpu": 33.02, "tokens/trainable": 129653} +{"epoch": 1.109375, "grad_norm": 0.39871567487716675, "learning_rate": 5.079691724810461e-05, "loss": 0.004961640574038029, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00497, "step": 284, "tokens/total": 8588384, "tokens/train_per_sec_per_gpu": 32.58, "tokens/trainable": 130100} +{"epoch": 1.11328125, "grad_norm": 0.3788130283355713, "learning_rate": 5.0507984812754684e-05, "loss": 0.009439138695597649, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00948, "step": 285, "tokens/total": 8618768, "tokens/train_per_sec_per_gpu": 30.48, "tokens/trainable": 130543} +{"epoch": 1.1171875, "grad_norm": 0.023167390376329422, "learning_rate": 5.021923930849237e-05, "loss": 0.00035375735023990273, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00035, "step": 286, "tokens/total": 8648976, "tokens/train_per_sec_per_gpu": 34.27, "tokens/trainable": 131000} +{"epoch": 1.12109375, "grad_norm": 0.010168107226490974, "learning_rate": 4.99306927511967e-05, "loss": 0.00027046652394346893, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00027, "step": 287, "tokens/total": 8679152, "tokens/train_per_sec_per_gpu": 34.04, "tokens/trainable": 131449} +{"epoch": 1.125, "grad_norm": 0.03659353405237198, "learning_rate": 4.964235714846775e-05, "loss": 0.0005067181773483753, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00051, "step": 288, "tokens/total": 8709504, "tokens/train_per_sec_per_gpu": 36.0, "tokens/trainable": 131923} +{"epoch": 1.12890625, "grad_norm": 0.03220707178115845, "learning_rate": 4.9354244499126866e-05, "loss": 0.000319857761496678, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00032, "step": 289, "tokens/total": 8739680, "tokens/train_per_sec_per_gpu": 33.03, "tokens/trainable": 132397} +{"epoch": 1.1328125, "grad_norm": 0.031606193631887436, "learning_rate": 4.90663667927174e-05, "loss": 0.0005622455500997603, "memory/device_reserved (GiB)": 35.57, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00056, "step": 290, "tokens/total": 8770112, "tokens/train_per_sec_per_gpu": 36.3, "tokens/trainable": 132846} +{"epoch": 1.13671875, "grad_norm": 0.6305046677589417, "learning_rate": 4.877873600900581e-05, "loss": 0.006522824056446552, "memory/device_reserved (GiB)": 35.57, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00654, "step": 291, "tokens/total": 8800368, "tokens/train_per_sec_per_gpu": 35.71, "tokens/trainable": 133347} +{"epoch": 1.140625, "grad_norm": 0.03814006596803665, "learning_rate": 4.849136411748306e-05, "loss": 0.000703076075296849, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.0007, "step": 292, "tokens/total": 8830688, "tokens/train_per_sec_per_gpu": 34.26, "tokens/trainable": 133779} +{"epoch": 1.14453125, "grad_norm": 0.05072004348039627, "learning_rate": 4.8204263076866574e-05, "loss": 0.0011388614075258374, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00114, "step": 293, "tokens/total": 8861008, "tokens/train_per_sec_per_gpu": 36.29, "tokens/trainable": 134251} +{"epoch": 1.1484375, "grad_norm": 0.2358824610710144, "learning_rate": 4.791744483460251e-05, "loss": 0.006398226134479046, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00642, "step": 294, "tokens/total": 8891760, "tokens/train_per_sec_per_gpu": 33.57, "tokens/trainable": 134683} +{"epoch": 1.15234375, "grad_norm": 0.2783265709877014, "learning_rate": 4.7630921326368736e-05, "loss": 0.0070259119383990765, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00705, "step": 295, "tokens/total": 8922192, "tokens/train_per_sec_per_gpu": 35.55, "tokens/trainable": 135165} +{"epoch": 1.15625, "grad_norm": 0.34820756316185, "learning_rate": 4.7344704475577916e-05, "loss": 0.004574563354253769, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00459, "step": 296, "tokens/total": 8952496, "tokens/train_per_sec_per_gpu": 33.23, "tokens/trainable": 135599} +{"epoch": 1.16015625, "grad_norm": 0.03794313222169876, "learning_rate": 4.705880619288153e-05, "loss": 0.0009930897504091263, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00099, "step": 297, "tokens/total": 8982768, "tokens/train_per_sec_per_gpu": 32.68, "tokens/trainable": 136038} +{"epoch": 1.1640625, "grad_norm": 0.19601035118103027, "learning_rate": 4.677323837567412e-05, "loss": 0.003028921550139785, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00303, "step": 298, "tokens/total": 9013248, "tokens/train_per_sec_per_gpu": 32.88, "tokens/trainable": 136495} +{"epoch": 1.16796875, "grad_norm": 0.0572715662419796, "learning_rate": 4.6488012907598146e-05, "loss": 0.001314631663262844, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00132, "step": 299, "tokens/total": 9043632, "tokens/train_per_sec_per_gpu": 35.29, "tokens/trainable": 136953} +{"epoch": 1.171875, "grad_norm": 0.12685413658618927, "learning_rate": 4.620314165804964e-05, "loss": 0.0023695232812315226, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00237, "step": 300, "tokens/total": 9073936, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 137412} +{"epoch": 1.17578125, "grad_norm": 0.04133208841085434, "learning_rate": 4.591863648168407e-05, "loss": 0.0008228939259424806, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00082, "step": 301, "tokens/total": 9104416, "tokens/train_per_sec_per_gpu": 35.68, "tokens/trainable": 137907} +{"epoch": 1.1796875, "grad_norm": 0.01651018112897873, "learning_rate": 4.5634509217923135e-05, "loss": 0.0003794343792833388, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00038, "step": 302, "tokens/total": 9134688, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 138383} +{"epoch": 1.18359375, "grad_norm": 0.009257322177290916, "learning_rate": 4.535077169046201e-05, "loss": 0.00024720095098018646, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00025, "step": 303, "tokens/total": 9165088, "tokens/train_per_sec_per_gpu": 36.42, "tokens/trainable": 138842} +{"epoch": 1.1875, "grad_norm": 0.038815200328826904, "learning_rate": 4.506743570677743e-05, "loss": 0.000565327238291502, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00057, "step": 304, "tokens/total": 9195584, "tokens/train_per_sec_per_gpu": 30.67, "tokens/trainable": 139269} +{"epoch": 1.19140625, "grad_norm": 0.37655678391456604, "learning_rate": 4.478451305763618e-05, "loss": 0.005295810755342245, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00531, "step": 305, "tokens/total": 9225760, "tokens/train_per_sec_per_gpu": 34.44, "tokens/trainable": 139736} +{"epoch": 1.1953125, "grad_norm": 0.014935888350009918, "learning_rate": 4.450201551660454e-05, "loss": 0.0003102960763499141, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00031, "step": 306, "tokens/total": 9256352, "tokens/train_per_sec_per_gpu": 37.0, "tokens/trainable": 140226} +{"epoch": 1.19921875, "grad_norm": 0.02618814818561077, "learning_rate": 4.4219954839558276e-05, "loss": 0.0003336220979690552, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.19, "memory/max_allocated (GiB)": 33.19, "ppl": 1.00033, "step": 307, "tokens/total": 9284256, "tokens/train_per_sec_per_gpu": 34.03, "tokens/trainable": 140664} +{"epoch": 1.203125, "grad_norm": 0.2529093027114868, "learning_rate": 4.393834276419352e-05, "loss": 0.005605834536254406, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00562, "step": 308, "tokens/total": 9314768, "tokens/train_per_sec_per_gpu": 29.95, "tokens/trainable": 141115} +{"epoch": 1.20703125, "grad_norm": 0.0079645412042737, "learning_rate": 4.36571910095382e-05, "loss": 0.00018522725440561771, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00019, "step": 309, "tokens/total": 9344976, "tokens/train_per_sec_per_gpu": 28.95, "tokens/trainable": 141530} +{"epoch": 1.2109375, "grad_norm": 0.4591776132583618, "learning_rate": 4.337651127546448e-05, "loss": 0.003637854475528002, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00364, "step": 310, "tokens/total": 9375280, "tokens/train_per_sec_per_gpu": 34.16, "tokens/trainable": 141964} +{"epoch": 1.21484375, "grad_norm": 0.02764446660876274, "learning_rate": 4.3096315242201736e-05, "loss": 0.00038255995605140924, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00038, "step": 311, "tokens/total": 9405536, "tokens/train_per_sec_per_gpu": 33.98, "tokens/trainable": 142428} +{"epoch": 1.21875, "grad_norm": 0.0073210448026657104, "learning_rate": 4.2816614569850635e-05, "loss": 0.0001625180011615157, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00016, "step": 312, "tokens/total": 9435696, "tokens/train_per_sec_per_gpu": 38.11, "tokens/trainable": 142894} +{"epoch": 1.22265625, "grad_norm": 0.1570003777742386, "learning_rate": 4.2537420897897864e-05, "loss": 0.00023247752687893808, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00023, "step": 313, "tokens/total": 9466080, "tokens/train_per_sec_per_gpu": 36.39, "tokens/trainable": 143368} +{"epoch": 1.2265625, "grad_norm": 0.2202703058719635, "learning_rate": 4.225874584473174e-05, "loss": 0.002779184840619564, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00278, "step": 314, "tokens/total": 9496400, "tokens/train_per_sec_per_gpu": 32.88, "tokens/trainable": 143824} +{"epoch": 1.23046875, "grad_norm": 0.04489271715283394, "learning_rate": 4.19806010071587e-05, "loss": 0.0008375659235753119, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00084, "step": 315, "tokens/total": 9526880, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 144306} +{"epoch": 1.234375, "grad_norm": 0.006173362489789724, "learning_rate": 4.170299795992081e-05, "loss": 0.0001250750501640141, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00013, "step": 316, "tokens/total": 9557280, "tokens/train_per_sec_per_gpu": 35.42, "tokens/trainable": 144763} +{"epoch": 1.23828125, "grad_norm": 0.011352705769240856, "learning_rate": 4.142594825521398e-05, "loss": 0.00022372312378138304, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00022, "step": 317, "tokens/total": 9587712, "tokens/train_per_sec_per_gpu": 37.46, "tokens/trainable": 145255} +{"epoch": 1.2421875, "grad_norm": 0.05640754476189613, "learning_rate": 4.114946342220728e-05, "loss": 0.0007214375073090196, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00072, "step": 318, "tokens/total": 9618064, "tokens/train_per_sec_per_gpu": 36.7, "tokens/trainable": 145714} +{"epoch": 1.24609375, "grad_norm": 0.15869443118572235, "learning_rate": 4.087355496656321e-05, "loss": 0.0053039658814668655, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00532, "step": 319, "tokens/total": 9648336, "tokens/train_per_sec_per_gpu": 36.04, "tokens/trainable": 146200} +{"epoch": 1.25, "grad_norm": 0.44812121987342834, "learning_rate": 4.05982343699588e-05, "loss": 0.0036084880121052265, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00362, "step": 320, "tokens/total": 9678688, "tokens/train_per_sec_per_gpu": 34.29, "tokens/trainable": 146668} +{"epoch": 1.25390625, "grad_norm": 0.05665228143334389, "learning_rate": 4.0323513089607876e-05, "loss": 0.0009465725161135197, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00095, "step": 321, "tokens/total": 9708848, "tokens/train_per_sec_per_gpu": 35.36, "tokens/trainable": 147114} +{"epoch": 1.2578125, "grad_norm": 0.016061583533883095, "learning_rate": 4.004940255778431e-05, "loss": 0.00015383574645966291, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00015, "step": 322, "tokens/total": 9739360, "tokens/train_per_sec_per_gpu": 35.2, "tokens/trainable": 147564} +{"epoch": 1.26171875, "grad_norm": 0.06347034126520157, "learning_rate": 3.977591418134619e-05, "loss": 0.0009517542785033584, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00095, "step": 323, "tokens/total": 9769776, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 148013} +{"epoch": 1.265625, "grad_norm": 0.24154019355773926, "learning_rate": 3.95030593412612e-05, "loss": 0.004087492823600769, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.0041, "step": 324, "tokens/total": 9800096, "tokens/train_per_sec_per_gpu": 38.07, "tokens/trainable": 148470} +{"epoch": 1.26953125, "grad_norm": 0.22498610615730286, "learning_rate": 3.923084939213296e-05, "loss": 0.0027766539715230465, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00278, "step": 325, "tokens/total": 9830304, "tokens/train_per_sec_per_gpu": 36.56, "tokens/trainable": 148926} +{"epoch": 1.2734375, "grad_norm": 0.002810331992805004, "learning_rate": 3.895929566172861e-05, "loss": 4.6118868340272456e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00005, "step": 326, "tokens/total": 9860608, "tokens/train_per_sec_per_gpu": 37.8, "tokens/trainable": 149420} +{"epoch": 1.27734375, "grad_norm": 0.007315841969102621, "learning_rate": 3.868840945050728e-05, "loss": 0.00011396597255952656, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00011, "step": 327, "tokens/total": 9890560, "tokens/train_per_sec_per_gpu": 39.43, "tokens/trainable": 149873} +{"epoch": 1.28125, "grad_norm": 0.021338002756237984, "learning_rate": 3.841820203114995e-05, "loss": 0.00021865982853341848, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00022, "step": 328, "tokens/total": 9920880, "tokens/train_per_sec_per_gpu": 33.84, "tokens/trainable": 150347} +{"epoch": 1.28515625, "grad_norm": 0.027939820662140846, "learning_rate": 3.814868464809027e-05, "loss": 0.0003473583492450416, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00035, "step": 329, "tokens/total": 9951280, "tokens/train_per_sec_per_gpu": 31.84, "tokens/trainable": 150822} +{"epoch": 1.2890625, "grad_norm": 0.0030000312253832817, "learning_rate": 3.787986851704667e-05, "loss": 5.043415876571089e-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": 9981600, "tokens/train_per_sec_per_gpu": 35.02, "tokens/trainable": 151266} +{"epoch": 1.29296875, "grad_norm": 0.016975272446870804, "learning_rate": 3.7611764824555654e-05, "loss": 0.0002860842505469918, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00029, "step": 331, "tokens/total": 10012016, "tokens/train_per_sec_per_gpu": 37.1, "tokens/trainable": 151734} +{"epoch": 1.296875, "grad_norm": 0.0012887469492852688, "learning_rate": 3.734438472750619e-05, "loss": 2.801024675136432e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00003, "step": 332, "tokens/total": 10042448, "tokens/train_per_sec_per_gpu": 33.99, "tokens/trainable": 152184} +{"epoch": 1.30078125, "grad_norm": 0.0035721107851713896, "learning_rate": 3.707773935267552e-05, "loss": 4.7658577386755496e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00005, "step": 333, "tokens/total": 10073024, "tokens/train_per_sec_per_gpu": 31.74, "tokens/trainable": 152623} +{"epoch": 1.3046875, "grad_norm": 0.036152660846710205, "learning_rate": 3.68118397962661e-05, "loss": 0.00040513710700906813, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00041, "step": 334, "tokens/total": 10103504, "tokens/train_per_sec_per_gpu": 33.4, "tokens/trainable": 153040} +{"epoch": 1.30859375, "grad_norm": 0.11260190606117249, "learning_rate": 3.654669712344384e-05, "loss": 0.0015010370407253504, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.31, "memory/max_allocated (GiB)": 33.31, "ppl": 1.0015, "step": 335, "tokens/total": 10131632, "tokens/train_per_sec_per_gpu": 31.76, "tokens/trainable": 153463} +{"epoch": 1.3125, "grad_norm": 0.17602311074733734, "learning_rate": 3.628232236787763e-05, "loss": 0.007916431874036789, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00795, "step": 336, "tokens/total": 10161824, "tokens/train_per_sec_per_gpu": 33.28, "tokens/trainable": 153908} +{"epoch": 1.31640625, "grad_norm": 0.005530365742743015, "learning_rate": 3.6018726531280144e-05, "loss": 8.388348942389712e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00008, "step": 337, "tokens/total": 10192448, "tokens/train_per_sec_per_gpu": 33.62, "tokens/trainable": 154387} +{"epoch": 1.3203125, "grad_norm": 0.06029028072953224, "learning_rate": 3.575592058295017e-05, "loss": 0.0002887388109229505, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00029, "step": 338, "tokens/total": 10222704, "tokens/train_per_sec_per_gpu": 33.94, "tokens/trainable": 154836} +{"epoch": 1.32421875, "grad_norm": 0.003610784187912941, "learning_rate": 3.549391545931585e-05, "loss": 6.314883648883551e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00006, "step": 339, "tokens/total": 10253168, "tokens/train_per_sec_per_gpu": 29.88, "tokens/trainable": 155261} +{"epoch": 1.328125, "grad_norm": 0.0029664942994713783, "learning_rate": 3.5232722063479914e-05, "loss": 4.0910730604082346e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00004, "step": 340, "tokens/total": 10283552, "tokens/train_per_sec_per_gpu": 31.18, "tokens/trainable": 155674} +{"epoch": 1.33203125, "grad_norm": 0.3301459848880768, "learning_rate": 3.49723512647657e-05, "loss": 0.001420877524651587, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00142, "step": 341, "tokens/total": 10313872, "tokens/train_per_sec_per_gpu": 35.76, "tokens/trainable": 156164} +{"epoch": 1.3359375, "grad_norm": 0.0065552121959626675, "learning_rate": 3.471281389826491e-05, "loss": 3.5097506042802706e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00004, "step": 342, "tokens/total": 10344256, "tokens/train_per_sec_per_gpu": 33.64, "tokens/trainable": 156611} +{"epoch": 1.33984375, "grad_norm": 0.027132108807563782, "learning_rate": 3.4454120764386764e-05, "loss": 0.00014500679390039295, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00015, "step": 343, "tokens/total": 10374544, "tokens/train_per_sec_per_gpu": 33.68, "tokens/trainable": 157073} +{"epoch": 1.34375, "grad_norm": 0.40091755986213684, "learning_rate": 3.4196282628408526e-05, "loss": 0.004070552997291088, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00408, "step": 344, "tokens/total": 10405040, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 157523} +{"epoch": 1.34765625, "grad_norm": 0.2770630419254303, "learning_rate": 3.3939310220027456e-05, "loss": 0.005590756889432669, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00561, "step": 345, "tokens/total": 10435424, "tokens/train_per_sec_per_gpu": 35.15, "tokens/trainable": 157964} +{"epoch": 1.3515625, "grad_norm": 0.007484348025172949, "learning_rate": 3.3683214232914404e-05, "loss": 8.245774370152503e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00008, "step": 346, "tokens/total": 10465760, "tokens/train_per_sec_per_gpu": 30.43, "tokens/trainable": 158390} +{"epoch": 1.35546875, "grad_norm": 0.005284447688609362, "learning_rate": 3.342800532426873e-05, "loss": 0.00010297571134287864, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.0001, "step": 347, "tokens/total": 10495904, "tokens/train_per_sec_per_gpu": 35.58, "tokens/trainable": 158848} +{"epoch": 1.359375, "grad_norm": 0.0784892812371254, "learning_rate": 3.317369411437484e-05, "loss": 0.0006088463123887777, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 34.0, "memory/max_allocated (GiB)": 34.0, "ppl": 1.00061, "step": 348, "tokens/total": 10526624, "tokens/train_per_sec_per_gpu": 28.7, "tokens/trainable": 159270} +{"epoch": 1.36328125, "grad_norm": 0.33535251021385193, "learning_rate": 3.292029118616024e-05, "loss": 0.006133331451565027, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00615, "step": 349, "tokens/total": 10556752, "tokens/train_per_sec_per_gpu": 32.5, "tokens/trainable": 159720} +{"epoch": 1.3671875, "grad_norm": 0.008170238696038723, "learning_rate": 3.266780708475511e-05, "loss": 0.00012147967208875343, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00012, "step": 350, "tokens/total": 10587264, "tokens/train_per_sec_per_gpu": 35.09, "tokens/trainable": 160187} +{"epoch": 1.37109375, "grad_norm": 0.001569622429087758, "learning_rate": 3.241625231705354e-05, "loss": 3.8519159716088325e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00004, "step": 351, "tokens/total": 10617504, "tokens/train_per_sec_per_gpu": 32.58, "tokens/trainable": 160645} +{"epoch": 1.375, "grad_norm": 0.143654003739357, "learning_rate": 3.216563735127618e-05, "loss": 0.002758543938398361, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00276, "step": 352, "tokens/total": 10647760, "tokens/train_per_sec_per_gpu": 34.45, "tokens/trainable": 161102} +{"epoch": 1.37890625, "grad_norm": 0.042433012276887894, "learning_rate": 3.191597261653475e-05, "loss": 0.00035650472273118794, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00036, "step": 353, "tokens/total": 10678080, "tokens/train_per_sec_per_gpu": 28.87, "tokens/trainable": 161555} +{"epoch": 1.3828125, "grad_norm": 0.9453999996185303, "learning_rate": 3.166726850239794e-05, "loss": 0.0030506744515150785, "memory/device_reserved (GiB)": 35.41, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00306, "step": 354, "tokens/total": 10708544, "tokens/train_per_sec_per_gpu": 36.67, "tokens/trainable": 162020} +{"epoch": 1.38671875, "grad_norm": 0.03272778168320656, "learning_rate": 3.141953535845912e-05, "loss": 0.00023931957548484206, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00024, "step": 355, "tokens/total": 10739024, "tokens/train_per_sec_per_gpu": 32.04, "tokens/trainable": 162448} +{"epoch": 1.390625, "grad_norm": 0.006854584906250238, "learning_rate": 3.11727834939056e-05, "loss": 6.610051059396937e-05, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.61, "memory/max_allocated (GiB)": 33.61, "ppl": 1.00007, "step": 356, "tokens/total": 10769024, "tokens/train_per_sec_per_gpu": 28.22, "tokens/trainable": 162863} +{"epoch": 1.39453125, "grad_norm": 0.03258929029107094, "learning_rate": 3.092702317708967e-05, "loss": 0.0002595578844193369, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00026, "step": 357, "tokens/total": 10799328, "tokens/train_per_sec_per_gpu": 34.94, "tokens/trainable": 163318} +{"epoch": 1.3984375, "grad_norm": 0.21498654782772064, "learning_rate": 3.0682264635101276e-05, "loss": 0.003778135636821389, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00379, "step": 358, "tokens/total": 10829488, "tokens/train_per_sec_per_gpu": 32.92, "tokens/trainable": 163767} +{"epoch": 1.40234375, "grad_norm": 0.22063465416431427, "learning_rate": 3.0438518053342407e-05, "loss": 0.00428510969504714, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00429, "step": 359, "tokens/total": 10859936, "tokens/train_per_sec_per_gpu": 31.9, "tokens/trainable": 164208} +{"epoch": 1.40625, "grad_norm": 0.01306198537349701, "learning_rate": 3.0195793575103266e-05, "loss": 0.00016752032388467342, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00017, "step": 360, "tokens/total": 10890400, "tokens/train_per_sec_per_gpu": 34.11, "tokens/trainable": 164655} +{"epoch": 1.41015625, "grad_norm": 0.42359107732772827, "learning_rate": 2.9954101301140146e-05, "loss": 0.022302038967609406, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02255, "step": 361, "tokens/total": 10920624, "tokens/train_per_sec_per_gpu": 34.09, "tokens/trainable": 165100} +{"epoch": 1.4140625, "grad_norm": 0.14819112420082092, "learning_rate": 2.9713451289255123e-05, "loss": 0.002135487971827388, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00214, "step": 362, "tokens/total": 10951136, "tokens/train_per_sec_per_gpu": 34.92, "tokens/trainable": 165587} +{"epoch": 1.41796875, "grad_norm": 0.10172182321548462, "learning_rate": 2.9473853553877484e-05, "loss": 0.0012491067172959447, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00125, "step": 363, "tokens/total": 10981344, "tokens/train_per_sec_per_gpu": 34.1, "tokens/trainable": 166034} +{"epoch": 1.421875, "grad_norm": 0.16598454117774963, "learning_rate": 2.9235318065647e-05, "loss": 0.0014915807405486703, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00149, "step": 364, "tokens/total": 11011552, "tokens/train_per_sec_per_gpu": 34.01, "tokens/trainable": 166485} +{"epoch": 1.42578125, "grad_norm": 0.01892480067908764, "learning_rate": 2.8997854750998964e-05, "loss": 0.00023381829669233412, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00023, "step": 365, "tokens/total": 11041872, "tokens/train_per_sec_per_gpu": 36.28, "tokens/trainable": 166973} +{"epoch": 1.4296875, "grad_norm": 0.050091467797756195, "learning_rate": 2.8761473491751258e-05, "loss": 0.00035884848330169916, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00036, "step": 366, "tokens/total": 11072192, "tokens/train_per_sec_per_gpu": 29.1, "tokens/trainable": 167414} +{"epoch": 1.43359375, "grad_norm": 0.6703732013702393, "learning_rate": 2.8526184124692883e-05, "loss": 0.008329304866492748, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00836, "step": 367, "tokens/total": 11102736, "tokens/train_per_sec_per_gpu": 30.18, "tokens/trainable": 167851} +{"epoch": 1.4375, "grad_norm": 0.10087089240550995, "learning_rate": 2.829199644117484e-05, "loss": 0.0014869305305182934, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00149, "step": 368, "tokens/total": 11133056, "tokens/train_per_sec_per_gpu": 35.53, "tokens/trainable": 168294} +{"epoch": 1.44140625, "grad_norm": 0.03148820623755455, "learning_rate": 2.8058920186702553e-05, "loss": 0.00043267637374810874, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00043, "step": 369, "tokens/total": 11163568, "tokens/train_per_sec_per_gpu": 32.17, "tokens/trainable": 168769} +{"epoch": 1.4453125, "grad_norm": 0.01531990710645914, "learning_rate": 2.782696506053033e-05, "loss": 0.00021343299886211753, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00021, "step": 370, "tokens/total": 11193936, "tokens/train_per_sec_per_gpu": 36.75, "tokens/trainable": 169270} +{"epoch": 1.44921875, "grad_norm": 0.006444540806114674, "learning_rate": 2.7596140715257824e-05, "loss": 0.00014627019118051976, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00015, "step": 371, "tokens/total": 11224480, "tokens/train_per_sec_per_gpu": 34.99, "tokens/trainable": 169711} +{"epoch": 1.453125, "grad_norm": 0.31427323818206787, "learning_rate": 2.7366456756428184e-05, "loss": 0.00462032575160265, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00463, "step": 372, "tokens/total": 11254912, "tokens/train_per_sec_per_gpu": 32.84, "tokens/trainable": 170185} +{"epoch": 1.45703125, "grad_norm": 0.013480601832270622, "learning_rate": 2.7137922742128486e-05, "loss": 0.000224081173655577, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00022, "step": 373, "tokens/total": 11285104, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 170650} +{"epoch": 1.4609375, "grad_norm": 0.023016467690467834, "learning_rate": 2.691054818259188e-05, "loss": 0.0001827064115786925, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00018, "step": 374, "tokens/total": 11315600, "tokens/train_per_sec_per_gpu": 31.93, "tokens/trainable": 171126} +{"epoch": 1.46484375, "grad_norm": 0.019638920202851295, "learning_rate": 2.6684342539801933e-05, "loss": 0.0005156319821253419, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00052, "step": 375, "tokens/total": 11345776, "tokens/train_per_sec_per_gpu": 35.58, "tokens/trainable": 171600} +{"epoch": 1.46875, "grad_norm": 0.03251553326845169, "learning_rate": 2.645931522709877e-05, "loss": 0.0004913917509838939, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00049, "step": 376, "tokens/total": 11376208, "tokens/train_per_sec_per_gpu": 28.1, "tokens/trainable": 172017} +{"epoch": 1.47265625, "grad_norm": 0.03556421399116516, "learning_rate": 2.6235475608787365e-05, "loss": 0.0005220613675191998, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00052, "step": 377, "tokens/total": 11406512, "tokens/train_per_sec_per_gpu": 35.97, "tokens/trainable": 172493} +{"epoch": 1.4765625, "grad_norm": 0.04202103987336159, "learning_rate": 2.6012832999747916e-05, "loss": 0.0004697944095823914, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00047, "step": 378, "tokens/total": 11436544, "tokens/train_per_sec_per_gpu": 34.36, "tokens/trainable": 172951} +{"epoch": 1.48046875, "grad_norm": 0.023767679929733276, "learning_rate": 2.579139666504821e-05, "loss": 0.0004130054439883679, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00041, "step": 379, "tokens/total": 11467008, "tokens/train_per_sec_per_gpu": 35.65, "tokens/trainable": 173435} +{"epoch": 1.484375, "grad_norm": 0.04106076434254646, "learning_rate": 2.557117581955798e-05, "loss": 0.00047382546472363174, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00047, "step": 380, "tokens/total": 11497184, "tokens/train_per_sec_per_gpu": 30.39, "tokens/trainable": 173900} +{"epoch": 1.48828125, "grad_norm": 0.15864700078964233, "learning_rate": 2.5352179627565532e-05, "loss": 0.003480169689282775, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00349, "step": 381, "tokens/total": 11527600, "tokens/train_per_sec_per_gpu": 34.05, "tokens/trainable": 174371} +{"epoch": 1.4921875, "grad_norm": 0.04146522656083107, "learning_rate": 2.5134417202396277e-05, "loss": 0.0005609758081845939, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00056, "step": 382, "tokens/total": 11557808, "tokens/train_per_sec_per_gpu": 32.74, "tokens/trainable": 174826} +{"epoch": 1.49609375, "grad_norm": 0.012171853333711624, "learning_rate": 2.491789760603361e-05, "loss": 0.00022788115893490613, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00023, "step": 383, "tokens/total": 11588352, "tokens/train_per_sec_per_gpu": 32.86, "tokens/trainable": 175276} +{"epoch": 1.5, "grad_norm": 0.08181510865688324, "learning_rate": 2.4702629848741764e-05, "loss": 0.0016442432533949614, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00165, "step": 384, "tokens/total": 11618640, "tokens/train_per_sec_per_gpu": 33.67, "tokens/trainable": 175720} +{"epoch": 1.50390625, "grad_norm": 0.10246019810438156, "learning_rate": 2.4488622888690785e-05, "loss": 0.0017187006305903196, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00172, "step": 385, "tokens/total": 11649072, "tokens/train_per_sec_per_gpu": 35.21, "tokens/trainable": 176203} +{"epoch": 1.5078125, "grad_norm": 0.2453320175409317, "learning_rate": 2.427588563158384e-05, "loss": 0.0038096809294074774, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00382, "step": 386, "tokens/total": 11679552, "tokens/train_per_sec_per_gpu": 34.34, "tokens/trainable": 176651} +{"epoch": 1.51171875, "grad_norm": 0.005247004795819521, "learning_rate": 2.406442693028651e-05, "loss": 0.00010425677464809269, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0001, "step": 387, "tokens/total": 11709760, "tokens/train_per_sec_per_gpu": 32.69, "tokens/trainable": 177115} +{"epoch": 1.515625, "grad_norm": 0.027557479217648506, "learning_rate": 2.3854255584458547e-05, "loss": 0.00036113697569817305, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00036, "step": 388, "tokens/total": 11740288, "tokens/train_per_sec_per_gpu": 36.08, "tokens/trainable": 177601} +{"epoch": 1.51953125, "grad_norm": 0.028131073340773582, "learning_rate": 2.3645380340187508e-05, "loss": 0.0005047487211413682, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0005, "step": 389, "tokens/total": 11770592, "tokens/train_per_sec_per_gpu": 32.96, "tokens/trainable": 178050} +{"epoch": 1.5234375, "grad_norm": 0.0033069071359932423, "learning_rate": 2.3437809889624914e-05, "loss": 7.658063259441406e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00008, "step": 390, "tokens/total": 11800752, "tokens/train_per_sec_per_gpu": 36.4, "tokens/trainable": 178531} +{"epoch": 1.52734375, "grad_norm": 0.028486015275120735, "learning_rate": 2.3231552870624487e-05, "loss": 0.0005519004189409316, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00055, "step": 391, "tokens/total": 11831360, "tokens/train_per_sec_per_gpu": 33.91, "tokens/trainable": 179000} +{"epoch": 1.53125, "grad_norm": 0.002584393136203289, "learning_rate": 2.3026617866382657e-05, "loss": 5.4768250265624374e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00005, "step": 392, "tokens/total": 11861552, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 179457} +{"epoch": 1.53515625, "grad_norm": 0.009344775229692459, "learning_rate": 2.2823013405081507e-05, "loss": 0.0002210603270214051, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00022, "step": 393, "tokens/total": 11891904, "tokens/train_per_sec_per_gpu": 32.74, "tokens/trainable": 179890} +{"epoch": 1.5390625, "grad_norm": 0.004884887952357531, "learning_rate": 2.2620747959533722e-05, "loss": 8.428651199210435e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00008, "step": 394, "tokens/total": 11922208, "tokens/train_per_sec_per_gpu": 33.87, "tokens/trainable": 180341} +{"epoch": 1.54296875, "grad_norm": 0.023302890360355377, "learning_rate": 2.2419829946830123e-05, "loss": 0.000332253344822675, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00033, "step": 395, "tokens/total": 11952672, "tokens/train_per_sec_per_gpu": 32.87, "tokens/trainable": 180762} +{"epoch": 1.546875, "grad_norm": 0.14256854355335236, "learning_rate": 2.2220267727989325e-05, "loss": 0.00286297220736742, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00287, "step": 396, "tokens/total": 11983088, "tokens/train_per_sec_per_gpu": 35.13, "tokens/trainable": 181226} +{"epoch": 1.55078125, "grad_norm": 0.00555575592443347, "learning_rate": 2.202206960760984e-05, "loss": 9.771026816451922e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0001, "step": 397, "tokens/total": 12013488, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 181714} +{"epoch": 1.5546875, "grad_norm": 0.007696876302361488, "learning_rate": 2.182524383352446e-05, "loss": 0.0001580697571625933, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00016, "step": 398, "tokens/total": 12043968, "tokens/train_per_sec_per_gpu": 35.44, "tokens/trainable": 182142} +{"epoch": 1.55859375, "grad_norm": 0.0037960875779390335, "learning_rate": 2.1629798596457056e-05, "loss": 7.467059913324192e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00007, "step": 399, "tokens/total": 12074160, "tokens/train_per_sec_per_gpu": 34.82, "tokens/trainable": 182604} +{"epoch": 1.5625, "grad_norm": 0.004514187108725309, "learning_rate": 2.1435742029681725e-05, "loss": 0.00011837383499369025, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.67, "memory/max_allocated (GiB)": 33.67, "ppl": 1.00012, "step": 400, "tokens/total": 12104352, "tokens/train_per_sec_per_gpu": 25.26, "tokens/trainable": 183011} +{"epoch": 1.56640625, "grad_norm": 0.09468130767345428, "learning_rate": 2.124308220868431e-05, "loss": 0.0007745609618723392, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00077, "step": 401, "tokens/total": 12134240, "tokens/train_per_sec_per_gpu": 31.75, "tokens/trainable": 183459} +{"epoch": 1.5703125, "grad_norm": 0.017941495403647423, "learning_rate": 2.105182715082638e-05, "loss": 0.00024286848201882094, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00024, "step": 402, "tokens/total": 12164480, "tokens/train_per_sec_per_gpu": 38.62, "tokens/trainable": 183959} +{"epoch": 1.57421875, "grad_norm": 0.005621996242552996, "learning_rate": 2.0861984815011552e-05, "loss": 0.00012180282647022977, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00012, "step": 403, "tokens/total": 12194640, "tokens/train_per_sec_per_gpu": 32.46, "tokens/trainable": 184431} +{"epoch": 1.578125, "grad_norm": 0.004712392576038837, "learning_rate": 2.0673563101354323e-05, "loss": 0.00010195293725701049, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0001, "step": 404, "tokens/total": 12224960, "tokens/train_per_sec_per_gpu": 31.34, "tokens/trainable": 184867} +{"epoch": 1.58203125, "grad_norm": 0.023734599351882935, "learning_rate": 2.0486569850851317e-05, "loss": 0.00040538967004977167, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00041, "step": 405, "tokens/total": 12255008, "tokens/train_per_sec_per_gpu": 29.33, "tokens/trainable": 185300} +{"epoch": 1.5859375, "grad_norm": 0.06495791673660278, "learning_rate": 2.0301012845054956e-05, "loss": 0.000537894491571933, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00054, "step": 406, "tokens/total": 12285248, "tokens/train_per_sec_per_gpu": 31.47, "tokens/trainable": 185754} +{"epoch": 1.58984375, "grad_norm": 0.003490478964522481, "learning_rate": 2.011689980574966e-05, "loss": 6.170629058033228e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00006, "step": 407, "tokens/total": 12315552, "tokens/train_per_sec_per_gpu": 36.78, "tokens/trainable": 186213} +{"epoch": 1.59375, "grad_norm": 0.005349478218704462, "learning_rate": 1.993423839463052e-05, "loss": 9.109645179705694e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00009, "step": 408, "tokens/total": 12345904, "tokens/train_per_sec_per_gpu": 32.78, "tokens/trainable": 186659} +{"epoch": 1.59765625, "grad_norm": 0.011329016648232937, "learning_rate": 1.975303621298445e-05, "loss": 0.00016253325156867504, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00016, "step": 409, "tokens/total": 12376336, "tokens/train_per_sec_per_gpu": 33.0, "tokens/trainable": 187083} +{"epoch": 1.6015625, "grad_norm": 0.019948504865169525, "learning_rate": 1.957330080137385e-05, "loss": 0.00013768397911917418, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00014, "step": 410, "tokens/total": 12406880, "tokens/train_per_sec_per_gpu": 34.91, "tokens/trainable": 187530} +{"epoch": 1.60546875, "grad_norm": 0.003809648798778653, "learning_rate": 1.9395039639322864e-05, "loss": 5.0881026254501194e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00005, "step": 411, "tokens/total": 12437088, "tokens/train_per_sec_per_gpu": 34.28, "tokens/trainable": 187972} +{"epoch": 1.609375, "grad_norm": 0.012643440626561642, "learning_rate": 1.9218260145006073e-05, "loss": 0.0001802633050829172, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00018, "step": 412, "tokens/total": 12465440, "tokens/train_per_sec_per_gpu": 39.92, "tokens/trainable": 188417} +{"epoch": 1.61328125, "grad_norm": 0.22178205847740173, "learning_rate": 1.904296967493982e-05, "loss": 0.011118180118501186, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01118, "step": 413, "tokens/total": 12495968, "tokens/train_per_sec_per_gpu": 31.49, "tokens/trainable": 188868} +{"epoch": 1.6171875, "grad_norm": 0.0006713962648063898, "learning_rate": 1.8869175523676064e-05, "loss": 1.6992760720313527e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00002, "step": 414, "tokens/total": 12526128, "tokens/train_per_sec_per_gpu": 36.9, "tokens/trainable": 189337} +{"epoch": 1.62109375, "grad_norm": 0.006867737043648958, "learning_rate": 1.869688492349885e-05, "loss": 0.00011090705083915964, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00011, "step": 415, "tokens/total": 12556256, "tokens/train_per_sec_per_gpu": 31.79, "tokens/trainable": 189750} +{"epoch": 1.625, "grad_norm": 0.016067150980234146, "learning_rate": 1.85261050441233e-05, "loss": 0.0001882174692582339, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00019, "step": 416, "tokens/total": 12586736, "tokens/train_per_sec_per_gpu": 33.98, "tokens/trainable": 190216} +{"epoch": 1.62890625, "grad_norm": 0.0036702004726976156, "learning_rate": 1.8356842992397304e-05, "loss": 4.983436519978568e-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": 12617168, "tokens/train_per_sec_per_gpu": 34.94, "tokens/trainable": 190708} +{"epoch": 1.6328125, "grad_norm": 0.3933418095111847, "learning_rate": 1.8189105812005714e-05, "loss": 0.0050767892971634865, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00509, "step": 418, "tokens/total": 12647680, "tokens/train_per_sec_per_gpu": 30.59, "tokens/trainable": 191126} +{"epoch": 1.63671875, "grad_norm": 0.04893051087856293, "learning_rate": 1.802290048317732e-05, "loss": 0.0004771970270667225, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00048, "step": 419, "tokens/total": 12677952, "tokens/train_per_sec_per_gpu": 30.28, "tokens/trainable": 191540} +{"epoch": 1.640625, "grad_norm": 0.03733966872096062, "learning_rate": 1.785823392239424e-05, "loss": 0.0004862284695263952, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00049, "step": 420, "tokens/total": 12708416, "tokens/train_per_sec_per_gpu": 36.67, "tokens/trainable": 192014} +{"epoch": 1.64453125, "grad_norm": 0.00783330574631691, "learning_rate": 1.7695112982104225e-05, "loss": 7.455339073203504e-05, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00007, "step": 421, "tokens/total": 12738640, "tokens/train_per_sec_per_gpu": 31.28, "tokens/trainable": 192463} +{"epoch": 1.6484375, "grad_norm": 0.052442099899053574, "learning_rate": 1.7533544450435433e-05, "loss": 0.0009662319207563996, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00097, "step": 422, "tokens/total": 12769232, "tokens/train_per_sec_per_gpu": 32.64, "tokens/trainable": 192918} +{"epoch": 1.65234375, "grad_norm": 0.024694403633475304, "learning_rate": 1.7373535050913946e-05, "loss": 0.00027854571817442775, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00028, "step": 423, "tokens/total": 12799648, "tokens/train_per_sec_per_gpu": 37.36, "tokens/trainable": 193425} +{"epoch": 1.65625, "grad_norm": 0.002282459754496813, "learning_rate": 1.721509144218405e-05, "loss": 5.6570435845060274e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00006, "step": 424, "tokens/total": 12829920, "tokens/train_per_sec_per_gpu": 35.18, "tokens/trainable": 193898} +{"epoch": 1.66015625, "grad_norm": 0.28251373767852783, "learning_rate": 1.705822021773101e-05, "loss": 0.010694734752178192, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01075, "step": 425, "tokens/total": 12860112, "tokens/train_per_sec_per_gpu": 29.22, "tokens/trainable": 194315} +{"epoch": 1.6640625, "grad_norm": 0.14038732647895813, "learning_rate": 1.69029279056068e-05, "loss": 0.0015529554802924395, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00155, "step": 426, "tokens/total": 12890320, "tokens/train_per_sec_per_gpu": 34.88, "tokens/trainable": 194796} +{"epoch": 1.66796875, "grad_norm": 0.0007188008166849613, "learning_rate": 1.6749220968158415e-05, "loss": 1.8845212252927013e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00002, "step": 427, "tokens/total": 12920656, "tokens/train_per_sec_per_gpu": 33.91, "tokens/trainable": 195262} +{"epoch": 1.671875, "grad_norm": 0.002983702579513192, "learning_rate": 1.659710580175893e-05, "loss": 5.061700358055532e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00005, "step": 428, "tokens/total": 12951040, "tokens/train_per_sec_per_gpu": 31.88, "tokens/trainable": 195680} +{"epoch": 1.67578125, "grad_norm": 0.00501316599547863, "learning_rate": 1.644658873654133e-05, "loss": 8.430491288891062e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00008, "step": 429, "tokens/total": 12981232, "tokens/train_per_sec_per_gpu": 32.83, "tokens/trainable": 196145} +{"epoch": 1.6796875, "grad_norm": 0.6195209622383118, "learning_rate": 1.629767603613508e-05, "loss": 0.0007395793218165636, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00074, "step": 430, "tokens/total": 13011616, "tokens/train_per_sec_per_gpu": 32.37, "tokens/trainable": 196626} +{"epoch": 1.68359375, "grad_norm": 0.32764554023742676, "learning_rate": 1.615037389740547e-05, "loss": 0.012379031628370285, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.01246, "step": 431, "tokens/total": 13042208, "tokens/train_per_sec_per_gpu": 31.08, "tokens/trainable": 197067} +{"epoch": 1.6875, "grad_norm": 0.28968432545661926, "learning_rate": 1.600468845019576e-05, "loss": 0.003729558549821377, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00374, "step": 432, "tokens/total": 13072800, "tokens/train_per_sec_per_gpu": 38.65, "tokens/trainable": 197565} +{"epoch": 1.69140625, "grad_norm": 0.001470765215344727, "learning_rate": 1.5860625757072092e-05, "loss": 3.539394310791977e-05, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00004, "step": 433, "tokens/total": 13103328, "tokens/train_per_sec_per_gpu": 32.91, "tokens/trainable": 198015} +{"epoch": 1.6953125, "grad_norm": 0.019142981618642807, "learning_rate": 1.571819181307116e-05, "loss": 0.00024474930251017213, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00024, "step": 434, "tokens/total": 13133472, "tokens/train_per_sec_per_gpu": 33.86, "tokens/trainable": 198442} +{"epoch": 1.69921875, "grad_norm": 0.009595178067684174, "learning_rate": 1.557739254545075e-05, "loss": 0.0001564246922498569, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00016, "step": 435, "tokens/total": 13164064, "tokens/train_per_sec_per_gpu": 34.34, "tokens/trainable": 198914} +{"epoch": 1.703125, "grad_norm": 0.22099021077156067, "learning_rate": 1.543823381344311e-05, "loss": 0.014420191757380962, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.01452, "step": 436, "tokens/total": 13194464, "tokens/train_per_sec_per_gpu": 33.38, "tokens/trainable": 199355} +{"epoch": 1.70703125, "grad_norm": 0.1727798581123352, "learning_rate": 1.5300721408011114e-05, "loss": 0.008166543208062649, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0082, "step": 437, "tokens/total": 13224992, "tokens/train_per_sec_per_gpu": 32.95, "tokens/trainable": 199830} +{"epoch": 1.7109375, "grad_norm": 0.015094809234142303, "learning_rate": 1.5164861051607254e-05, "loss": 0.0002187314530601725, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00022, "step": 438, "tokens/total": 13255296, "tokens/train_per_sec_per_gpu": 34.88, "tokens/trainable": 200338} +{"epoch": 1.71484375, "grad_norm": 0.07529865950345993, "learning_rate": 1.5030658397935521e-05, "loss": 0.0006286898860707879, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00063, "step": 439, "tokens/total": 13285344, "tokens/train_per_sec_per_gpu": 32.81, "tokens/trainable": 200785} +{"epoch": 1.71875, "grad_norm": 0.005869010463356972, "learning_rate": 1.4898119031716104e-05, "loss": 0.00010347182978875935, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0001, "step": 440, "tokens/total": 13315632, "tokens/train_per_sec_per_gpu": 36.25, "tokens/trainable": 201267} +{"epoch": 1.72265625, "grad_norm": 0.02751794271171093, "learning_rate": 1.476724846845306e-05, "loss": 0.000567559851333499, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00057, "step": 441, "tokens/total": 13346048, "tokens/train_per_sec_per_gpu": 34.14, "tokens/trainable": 201734} +{"epoch": 1.7265625, "grad_norm": 0.06305643171072006, "learning_rate": 1.463805215420471e-05, "loss": 0.0012271346058696508, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00123, "step": 442, "tokens/total": 13376304, "tokens/train_per_sec_per_gpu": 30.3, "tokens/trainable": 202174} +{"epoch": 1.73046875, "grad_norm": 0.008338473737239838, "learning_rate": 1.451053546535705e-05, "loss": 0.00016968630370683968, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00017, "step": 443, "tokens/total": 13406720, "tokens/train_per_sec_per_gpu": 36.21, "tokens/trainable": 202612} +{"epoch": 1.734375, "grad_norm": 0.020954634994268417, "learning_rate": 1.438470370840001e-05, "loss": 0.0005087319877929986, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.64, "memory/max_allocated (GiB)": 33.64, "ppl": 1.00051, "step": 444, "tokens/total": 13436464, "tokens/train_per_sec_per_gpu": 31.0, "tokens/trainable": 203020} +{"epoch": 1.73828125, "grad_norm": 0.019224021583795547, "learning_rate": 1.4260562119706606e-05, "loss": 0.0003236978664062917, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00032, "step": 445, "tokens/total": 13466672, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 203479} +{"epoch": 1.7421875, "grad_norm": 0.02892594039440155, "learning_rate": 1.413811586531508e-05, "loss": 0.0004480727366171777, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00045, "step": 446, "tokens/total": 13496736, "tokens/train_per_sec_per_gpu": 35.31, "tokens/trainable": 203946} +{"epoch": 1.74609375, "grad_norm": 0.013984446413815022, "learning_rate": 1.4017370040713884e-05, "loss": 0.00018316772184334695, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00018, "step": 447, "tokens/total": 13526928, "tokens/train_per_sec_per_gpu": 37.45, "tokens/trainable": 204382} +{"epoch": 1.75, "grad_norm": 0.009344048798084259, "learning_rate": 1.3898329670629645e-05, "loss": 0.00011864845873787999, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00012, "step": 448, "tokens/total": 13557392, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 204830} +{"epoch": 1.75390625, "grad_norm": 0.022271448746323586, "learning_rate": 1.3780999708818058e-05, "loss": 0.00033031520433723927, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00033, "step": 449, "tokens/total": 13587776, "tokens/train_per_sec_per_gpu": 33.66, "tokens/trainable": 205300} +{"epoch": 1.7578125, "grad_norm": 0.3953968286514282, "learning_rate": 1.3665385037857758e-05, "loss": 0.0029637760017067194, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00297, "step": 450, "tokens/total": 13618112, "tokens/train_per_sec_per_gpu": 30.65, "tokens/trainable": 205726} +{"epoch": 1.76171875, "grad_norm": 0.008145474828779697, "learning_rate": 1.3551490468947126e-05, "loss": 0.00017418013885617256, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00017, "step": 451, "tokens/total": 13648512, "tokens/train_per_sec_per_gpu": 31.66, "tokens/trainable": 206163} +{"epoch": 1.765625, "grad_norm": 0.18700900673866272, "learning_rate": 1.3439320741704075e-05, "loss": 0.011147091165184975, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01121, "step": 452, "tokens/total": 13678704, "tokens/train_per_sec_per_gpu": 35.39, "tokens/trainable": 206619} +{"epoch": 1.76953125, "grad_norm": 0.2209583818912506, "learning_rate": 1.3328880523968808e-05, "loss": 0.004581573884934187, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00459, "step": 453, "tokens/total": 13709232, "tokens/train_per_sec_per_gpu": 37.09, "tokens/trainable": 207101} +{"epoch": 1.7734375, "grad_norm": 0.04667079448699951, "learning_rate": 1.3220174411609587e-05, "loss": 0.0004827629018109292, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00048, "step": 454, "tokens/total": 13739600, "tokens/train_per_sec_per_gpu": 35.7, "tokens/trainable": 207546} +{"epoch": 1.77734375, "grad_norm": 0.22248415648937225, "learning_rate": 1.3113206928331471e-05, "loss": 0.003234278876334429, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00324, "step": 455, "tokens/total": 13769936, "tokens/train_per_sec_per_gpu": 34.28, "tokens/trainable": 207989} +{"epoch": 1.78125, "grad_norm": 0.006573830731213093, "learning_rate": 1.300798252548806e-05, "loss": 0.00013498595217242837, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00013, "step": 456, "tokens/total": 13800240, "tokens/train_per_sec_per_gpu": 33.91, "tokens/trainable": 208402} +{"epoch": 1.78515625, "grad_norm": 0.0018037718255072832, "learning_rate": 1.2904505581896265e-05, "loss": 3.7187150155659765e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00004, "step": 457, "tokens/total": 13830512, "tokens/train_per_sec_per_gpu": 36.63, "tokens/trainable": 208904} +{"epoch": 1.7890625, "grad_norm": 0.01422063261270523, "learning_rate": 1.2802780403654082e-05, "loss": 0.00024471062351949513, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00024, "step": 458, "tokens/total": 13860832, "tokens/train_per_sec_per_gpu": 33.7, "tokens/trainable": 209382} +{"epoch": 1.79296875, "grad_norm": 0.01398254744708538, "learning_rate": 1.2702811223961408e-05, "loss": 0.000272503646556288, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00027, "step": 459, "tokens/total": 13890992, "tokens/train_per_sec_per_gpu": 33.87, "tokens/trainable": 209833} +{"epoch": 1.796875, "grad_norm": 0.12035191804170609, "learning_rate": 1.2604602202943861e-05, "loss": 0.0023982198908925056, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.0024, "step": 460, "tokens/total": 13921424, "tokens/train_per_sec_per_gpu": 32.49, "tokens/trainable": 210284} +{"epoch": 1.80078125, "grad_norm": 0.025684842839837074, "learning_rate": 1.2508157427479686e-05, "loss": 0.0005578990676440299, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.00056, "step": 461, "tokens/total": 13951504, "tokens/train_per_sec_per_gpu": 34.08, "tokens/trainable": 210769} +{"epoch": 1.8046875, "grad_norm": 0.0697178840637207, "learning_rate": 1.2413480911029655e-05, "loss": 0.0008336097234860063, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.38, "memory/max_allocated (GiB)": 33.38, "ppl": 1.00083, "step": 462, "tokens/total": 13979728, "tokens/train_per_sec_per_gpu": 30.42, "tokens/trainable": 211193} +{"epoch": 1.80859375, "grad_norm": 0.0039046104066073895, "learning_rate": 1.2320576593470082e-05, "loss": 9.185953240375966e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00009, "step": 463, "tokens/total": 14009936, "tokens/train_per_sec_per_gpu": 35.14, "tokens/trainable": 211655} +{"epoch": 1.8125, "grad_norm": 0.007744460366666317, "learning_rate": 1.2229448340928828e-05, "loss": 0.00017765363736543804, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00018, "step": 464, "tokens/total": 14039872, "tokens/train_per_sec_per_gpu": 29.3, "tokens/trainable": 212085} +{"epoch": 1.81640625, "grad_norm": 0.03343091905117035, "learning_rate": 1.2140099945624458e-05, "loss": 0.00046598759945482016, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00047, "step": 465, "tokens/total": 14070096, "tokens/train_per_sec_per_gpu": 32.67, "tokens/trainable": 212570} +{"epoch": 1.8203125, "grad_norm": 0.48290523886680603, "learning_rate": 1.205253512570841e-05, "loss": 0.016730966046452522, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01687, "step": 466, "tokens/total": 14100192, "tokens/train_per_sec_per_gpu": 32.9, "tokens/trainable": 213011} +{"epoch": 1.82421875, "grad_norm": 0.05692508816719055, "learning_rate": 1.1966757525110255e-05, "loss": 0.0009399584960192442, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00094, "step": 467, "tokens/total": 14130448, "tokens/train_per_sec_per_gpu": 31.58, "tokens/trainable": 213450} +{"epoch": 1.828125, "grad_norm": 0.008330056443810463, "learning_rate": 1.1882770713386095e-05, "loss": 0.0001491767616244033, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00015, "step": 468, "tokens/total": 14160480, "tokens/train_per_sec_per_gpu": 35.62, "tokens/trainable": 213898} +{"epoch": 1.83203125, "grad_norm": 0.005805026274174452, "learning_rate": 1.180057818556998e-05, "loss": 9.818439139053226e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0001, "step": 469, "tokens/total": 14191152, "tokens/train_per_sec_per_gpu": 36.85, "tokens/trainable": 214366} +{"epoch": 1.8359375, "grad_norm": 0.005398476962000132, "learning_rate": 1.1720183362028494e-05, "loss": 0.00013144969125278294, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00013, "step": 470, "tokens/total": 14219392, "tokens/train_per_sec_per_gpu": 35.72, "tokens/trainable": 214800} +{"epoch": 1.83984375, "grad_norm": 0.006290988065302372, "learning_rate": 1.1641589588318387e-05, "loss": 0.00014237320283427835, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00014, "step": 471, "tokens/total": 14249920, "tokens/train_per_sec_per_gpu": 35.44, "tokens/trainable": 215292} +{"epoch": 1.84375, "grad_norm": 0.0057176705449819565, "learning_rate": 1.1564800135047418e-05, "loss": 0.00012321076064836234, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00012, "step": 472, "tokens/total": 14280048, "tokens/train_per_sec_per_gpu": 33.89, "tokens/trainable": 215741} +{"epoch": 1.84765625, "grad_norm": 0.03017423115670681, "learning_rate": 1.148981819773816e-05, "loss": 0.0006377531681209803, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00064, "step": 473, "tokens/total": 14310432, "tokens/train_per_sec_per_gpu": 34.63, "tokens/trainable": 216194} +{"epoch": 1.8515625, "grad_norm": 0.04882626235485077, "learning_rate": 1.1416646896695086e-05, "loss": 0.0009199330816045403, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00092, "step": 474, "tokens/total": 14340496, "tokens/train_per_sec_per_gpu": 33.32, "tokens/trainable": 216634} +{"epoch": 1.85546875, "grad_norm": 0.49993258714675903, "learning_rate": 1.1345289276874717e-05, "loss": 0.0011110955383628607, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00111, "step": 475, "tokens/total": 14370720, "tokens/train_per_sec_per_gpu": 30.46, "tokens/trainable": 217040} +{"epoch": 1.859375, "grad_norm": 0.015277047641575336, "learning_rate": 1.1275748307758873e-05, "loss": 0.0002883652050513774, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00029, "step": 476, "tokens/total": 14401136, "tokens/train_per_sec_per_gpu": 37.65, "tokens/trainable": 217530} +{"epoch": 1.86328125, "grad_norm": 0.05111508071422577, "learning_rate": 1.1208026883231147e-05, "loss": 0.0007504059467464685, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00075, "step": 477, "tokens/total": 14431360, "tokens/train_per_sec_per_gpu": 34.43, "tokens/trainable": 217988} +{"epoch": 1.8671875, "grad_norm": 0.11643723398447037, "learning_rate": 1.1142127821456433e-05, "loss": 0.0013695022789761424, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00137, "step": 478, "tokens/total": 14461968, "tokens/train_per_sec_per_gpu": 32.58, "tokens/trainable": 218461} +{"epoch": 1.87109375, "grad_norm": 0.027246320620179176, "learning_rate": 1.1078053864763674e-05, "loss": 0.0004469662962947041, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00045, "step": 479, "tokens/total": 14491984, "tokens/train_per_sec_per_gpu": 32.81, "tokens/trainable": 218909} +{"epoch": 1.875, "grad_norm": 0.02049543894827366, "learning_rate": 1.1015807679531756e-05, "loss": 0.00044887623516842723, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00045, "step": 480, "tokens/total": 14522544, "tokens/train_per_sec_per_gpu": 34.06, "tokens/trainable": 219350} +{"epoch": 1.87890625, "grad_norm": 0.006593909580260515, "learning_rate": 1.0955391856078528e-05, "loss": 8.711963891983032e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00009, "step": 481, "tokens/total": 14552560, "tokens/train_per_sec_per_gpu": 29.5, "tokens/trainable": 219776} +{"epoch": 1.8828125, "grad_norm": 0.008904147893190384, "learning_rate": 1.0896808908553007e-05, "loss": 0.0001434069126844406, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00014, "step": 482, "tokens/total": 14583056, "tokens/train_per_sec_per_gpu": 35.08, "tokens/trainable": 220284} +{"epoch": 1.88671875, "grad_norm": 0.338387131690979, "learning_rate": 1.0840061274830763e-05, "loss": 0.004942934960126877, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00496, "step": 483, "tokens/total": 14613152, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 220714} +{"epoch": 1.890625, "grad_norm": 0.023702070116996765, "learning_rate": 1.0785151316412473e-05, "loss": 0.00027303973911330104, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00027, "step": 484, "tokens/total": 14643328, "tokens/train_per_sec_per_gpu": 35.03, "tokens/trainable": 221156} +{"epoch": 1.89453125, "grad_norm": 0.010423756204545498, "learning_rate": 1.0732081318325639e-05, "loss": 0.00012198913464089856, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00012, "step": 485, "tokens/total": 14673488, "tokens/train_per_sec_per_gpu": 31.64, "tokens/trainable": 221605} +{"epoch": 1.8984375, "grad_norm": 0.009971505962312222, "learning_rate": 1.0680853489029501e-05, "loss": 0.00017546328308526427, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00018, "step": 486, "tokens/total": 14704000, "tokens/train_per_sec_per_gpu": 30.45, "tokens/trainable": 222042} +{"epoch": 1.90234375, "grad_norm": 0.03453937545418739, "learning_rate": 1.0631469960323152e-05, "loss": 0.0005307079991325736, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.67, "memory/max_allocated (GiB)": 33.67, "ppl": 1.00053, "step": 487, "tokens/total": 14733984, "tokens/train_per_sec_per_gpu": 31.99, "tokens/trainable": 222474} +{"epoch": 1.90625, "grad_norm": 0.0030761805828660727, "learning_rate": 1.0583932787256783e-05, "loss": 5.944438453298062e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00006, "step": 488, "tokens/total": 14764400, "tokens/train_per_sec_per_gpu": 36.3, "tokens/trainable": 222967} +{"epoch": 1.91015625, "grad_norm": 0.008368371985852718, "learning_rate": 1.0538243948046206e-05, "loss": 0.00017430295702069998, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00017, "step": 489, "tokens/total": 14794672, "tokens/train_per_sec_per_gpu": 31.83, "tokens/trainable": 223420} +{"epoch": 1.9140625, "grad_norm": 0.0066870879381895065, "learning_rate": 1.0494405343990523e-05, "loss": 0.00010334253602195531, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.0001, "step": 490, "tokens/total": 14824784, "tokens/train_per_sec_per_gpu": 29.44, "tokens/trainable": 223864} +{"epoch": 1.91796875, "grad_norm": 0.0027329139411449432, "learning_rate": 1.0452418799392985e-05, "loss": 7.935120083857328e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00008, "step": 491, "tokens/total": 14855056, "tokens/train_per_sec_per_gpu": 32.88, "tokens/trainable": 224321} +{"epoch": 1.921875, "grad_norm": 0.005254871677607298, "learning_rate": 1.0412286061485102e-05, "loss": 9.15893615456298e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.00009, "step": 492, "tokens/total": 14885392, "tokens/train_per_sec_per_gpu": 36.35, "tokens/trainable": 224802} +{"epoch": 1.92578125, "grad_norm": 0.005493934731930494, "learning_rate": 1.03740088003539e-05, "loss": 0.00012532950495369732, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00013, "step": 493, "tokens/total": 14915760, "tokens/train_per_sec_per_gpu": 34.8, "tokens/trainable": 225284} +{"epoch": 1.9296875, "grad_norm": 0.01612873002886772, "learning_rate": 1.0337588608872463e-05, "loss": 0.00034530431730672717, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00035, "step": 494, "tokens/total": 14946192, "tokens/train_per_sec_per_gpu": 32.78, "tokens/trainable": 225733} +{"epoch": 1.93359375, "grad_norm": 0.0023398238699883223, "learning_rate": 1.0303027002633622e-05, "loss": 4.5659828174393624e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00005, "step": 495, "tokens/total": 14976272, "tokens/train_per_sec_per_gpu": 32.55, "tokens/trainable": 226172} +{"epoch": 1.9375, "grad_norm": 0.0058130365796387196, "learning_rate": 1.0270325419886884e-05, "loss": 0.0001502669183537364, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00015, "step": 496, "tokens/total": 15006544, "tokens/train_per_sec_per_gpu": 34.81, "tokens/trainable": 226636} +{"epoch": 1.94140625, "grad_norm": 0.044661372900009155, "learning_rate": 1.0239485221478599e-05, "loss": 0.0003950877580791712, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0004, "step": 497, "tokens/total": 15037040, "tokens/train_per_sec_per_gpu": 32.91, "tokens/trainable": 227067} +{"epoch": 1.9453125, "grad_norm": 0.05208620801568031, "learning_rate": 1.0210507690795292e-05, "loss": 0.0004586712457239628, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00046, "step": 498, "tokens/total": 15067344, "tokens/train_per_sec_per_gpu": 37.58, "tokens/trainable": 227532} +{"epoch": 1.94921875, "grad_norm": 0.058274783194065094, "learning_rate": 1.0183394033710305e-05, "loss": 0.0008681797189638019, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00087, "step": 499, "tokens/total": 15097648, "tokens/train_per_sec_per_gpu": 36.52, "tokens/trainable": 228007} +{"epoch": 1.953125, "grad_norm": 0.007533176336437464, "learning_rate": 1.0158145378533583e-05, "loss": 0.0001981634268304333, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.0002, "step": 500, "tokens/total": 15128000, "tokens/train_per_sec_per_gpu": 30.86, "tokens/trainable": 228457} +{"epoch": 1.95703125, "grad_norm": 0.007274497300386429, "learning_rate": 1.0134762775964726e-05, "loss": 0.00011473346239654347, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00011, "step": 501, "tokens/total": 15158352, "tokens/train_per_sec_per_gpu": 35.2, "tokens/trainable": 228928} +{"epoch": 1.9609375, "grad_norm": 0.008716201409697533, "learning_rate": 1.0113247199049278e-05, "loss": 0.00016994940233416855, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00017, "step": 502, "tokens/total": 15189136, "tokens/train_per_sec_per_gpu": 32.14, "tokens/trainable": 229368} +{"epoch": 1.96484375, "grad_norm": 0.04848707467317581, "learning_rate": 1.0093599543138205e-05, "loss": 0.0008516436209902167, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00085, "step": 503, "tokens/total": 15219616, "tokens/train_per_sec_per_gpu": 32.7, "tokens/trainable": 229815} +{"epoch": 1.96875, "grad_norm": 0.01895209774374962, "learning_rate": 1.0075820625850675e-05, "loss": 0.00022323444136418402, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00022, "step": 504, "tokens/total": 15250032, "tokens/train_per_sec_per_gpu": 38.87, "tokens/trainable": 230319} +{"epoch": 1.97265625, "grad_norm": 0.1530815213918686, "learning_rate": 1.0059911187040013e-05, "loss": 0.001839508069679141, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00184, "step": 505, "tokens/total": 15280592, "tokens/train_per_sec_per_gpu": 37.49, "tokens/trainable": 230807} +{"epoch": 1.9765625, "grad_norm": 0.049176666885614395, "learning_rate": 1.0045871888762893e-05, "loss": 0.00043575733434408903, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00044, "step": 506, "tokens/total": 15310816, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 231257} +{"epoch": 1.98046875, "grad_norm": 0.019041752442717552, "learning_rate": 1.003370331525184e-05, "loss": 0.00026456083287484944, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00026, "step": 507, "tokens/total": 15341424, "tokens/train_per_sec_per_gpu": 36.32, "tokens/trainable": 231746} +{"epoch": 1.984375, "grad_norm": 0.004318088758736849, "learning_rate": 1.002340597289085e-05, "loss": 6.76921772537753e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.33, "memory/max_allocated (GiB)": 33.33, "ppl": 1.00007, "step": 508, "tokens/total": 15369664, "tokens/train_per_sec_per_gpu": 33.0, "tokens/trainable": 232172} +{"epoch": 1.98828125, "grad_norm": 0.0025166042614728212, "learning_rate": 1.0014980290194387e-05, "loss": 4.969122164766304e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.00005, "step": 509, "tokens/total": 15400496, "tokens/train_per_sec_per_gpu": 37.11, "tokens/trainable": 232680} +{"epoch": 1.9921875, "grad_norm": 0.07566080242395401, "learning_rate": 1.0008426617789489e-05, "loss": 0.0005519739352166653, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00055, "step": 510, "tokens/total": 15430864, "tokens/train_per_sec_per_gpu": 36.04, "tokens/trainable": 233169} +{"epoch": 1.99609375, "grad_norm": 0.022425467148423195, "learning_rate": 1.0003745228401215e-05, "loss": 0.00015870542847551405, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00016, "step": 511, "tokens/total": 15461232, "tokens/train_per_sec_per_gpu": 32.72, "tokens/trainable": 233609} +{"epoch": 2.0, "grad_norm": 0.013609496876597404, "learning_rate": 1.0000936316841296e-05, "loss": 0.00012728071305900812, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00013, "step": 512, "tokens/total": 15491760, "tokens/train_per_sec_per_gpu": 36.19, "tokens/trainable": 234078}