diff --git a/.gitattributes b/.gitattributes index 67bda838058a831a0e2c036a70302ecb5e4e7ac5..a1a57dc774512c39b9e9a56b4a2cc68341e05099 100644 --- a/.gitattributes +++ b/.gitattributes @@ -494,3 +494,20 @@ aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-512/tokenizer.js aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/coin_real_4x__coin2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_real_1x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text diff --git a/aft_wave_v2/coin_real_1x__agreement/training/ARTIFACT_MANIFEST.local.json b/aft_wave_v2/coin_real_1x__agreement/training/ARTIFACT_MANIFEST.local.json new file mode 100644 index 0000000000000000000000000000000000000000..64b1fde52d31575ba097cf8dbabe8071071e83e8 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/ARTIFACT_MANIFEST.local.json @@ -0,0 +1,867 @@ +{ + "repo": "arcadia-impact/scimt-dispatch-models", + "remote_prefix": "aft_wave_v2/coin_real_1x__agreement/training", + "local_folder": "/workspace/wave/training", + "files": { + "TRAINED.json": { + "size": 957, + "sha256": "ed406492d51b88bd3d274e86fbd183a7785ac417babffe767a4900f9663e7133" + }, + "axolotl.yaml": { + "size": 1211, + "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e" + }, + "checkpoint.json": { + "size": 2224, + "sha256": "722676bed29a82b526bc75e0b1737644944cbbc23a25c3c7747120ae202624cf" + }, + "checkpoints/README.md": { + "size": 2937, + "sha256": "a5ac69b85486d95a8673c2de6b2e116a6b5069fdf10abbbe7ccb4b8d7f2e7317" + }, + "checkpoints/adapter_config.json": { + "size": 1098, + "sha256": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/adapter_model.safetensors": { + "size": 547777976, + "sha256": "82e5c558fd3518967c97a463cc23363818705003444ae831b5667d43e013cc03" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-128/adapter_model.safetensors": { + "size": 547777976, + "sha256": "9d6f7dc703af7a80756c2db7f916e936ba8f3fb1888e9774543777019deda638" + }, + "checkpoints/checkpoint-128/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-128/optimizer.pt": { + "size": 1048106435, + "sha256": "368b3dac3ce73c64c978a430804a3a64d276128c43f4b8d6f31d0f5379cb36e8" + }, + "checkpoints/checkpoint-128/rng_state.pth": { + "size": 14645, + "sha256": "9983d79410c5ae3ad4dd6b33d87b229e318829b73aa2e270784b5be7d242f714" + }, + "checkpoints/checkpoint-128/scheduler.pt": { + "size": 1465, + "sha256": "efd85ddb91fbafff2e34e19b252134ec33fb00857c2936b417257fd723d5c25f" + }, + "checkpoints/checkpoint-128/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-128/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-128/tokens_state.json": { + "size": 38, + "sha256": "a5dd8d174e00d61665d74bc66209b6461b1184f727a0d74174997597bb8b38a6" + }, + "checkpoints/checkpoint-128/trainer_state.json": { + "size": 56220, + "sha256": "fb4068c8a083f1882e131c46751df0be8632e5a4a5a24a6532c3f7646da2c912" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-160/adapter_model.safetensors": { + "size": 547777976, + "sha256": "fab5da99ffb1e531fe46e097946bd2e7cfe29250a2377f2e028f7034be1388c0" + }, + "checkpoints/checkpoint-160/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-160/optimizer.pt": { + "size": 1048106435, + "sha256": "62dfc9938a2107ed60a481bb92f478c5b103d1cb7499a5be2252d6378799c68c" + }, + "checkpoints/checkpoint-160/rng_state.pth": { + "size": 14645, + "sha256": "bbe19eb8e9b9385d32e98b493727887e743dd30947d7e742ac3b5158b9819536" + }, + "checkpoints/checkpoint-160/scheduler.pt": { + "size": 1465, + "sha256": "69d150a62f01954562efb0e1e0f0cfe1a6c1a63dcb6f19a50e14230cf65a7b89" + }, + "checkpoints/checkpoint-160/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-160/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-160/tokens_state.json": { + "size": 38, + "sha256": "a9c33caced7d456d62f444f1f50d8ffdf898b5a0e31c26d17c2f3edd05e45f64" + }, + "checkpoints/checkpoint-160/trainer_state.json": { + "size": 70202, + "sha256": "43c9d4168a770d9e4dc7eb30d61bf8b176ef10925517409cffa25f7f838f7403" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-192/adapter_model.safetensors": { + "size": 547777976, + "sha256": "a56c1fbd3eb063bfa0ce7a15a20577569c29fcec60eb4c20da60cbdab95b19dc" + }, + "checkpoints/checkpoint-192/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-192/optimizer.pt": { + "size": 1048106435, + "sha256": "8c2141d2e5550443cbf4a5cdf9f0cfb92cc5cd62fd8cf95f4c09d3fc8a321fdd" + }, + "checkpoints/checkpoint-192/rng_state.pth": { + "size": 14645, + "sha256": "0d896644f2e98c81e623938a072f96d9cb9eb180436ba03e68153edafca3bfcb" + }, + "checkpoints/checkpoint-192/scheduler.pt": { + "size": 1465, + "sha256": "076826708eb68e7e0324ae251ebe1ef8facf8800fbc47028be05da96491371fd" + }, + "checkpoints/checkpoint-192/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-192/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-192/tokens_state.json": { + "size": 38, + "sha256": "6dcb03db82c34201f9966e9d9c444c1957b95cd53c5ea9ed3b950f223ea74fbe" + }, + "checkpoints/checkpoint-192/trainer_state.json": { + "size": 84204, + "sha256": "36cb13a3002efb3cb7b02f2fbb1764e6dc963b11a9277d644ae7bdb4b0b8b078" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-224/adapter_model.safetensors": { + "size": 547777976, + "sha256": "26345eeb816e18811525163b2b729f6c6d1d7c1cd6d604d373815998d830ab55" + }, + "checkpoints/checkpoint-224/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-224/optimizer.pt": { + "size": 1048106435, + "sha256": "8970aa983a72b53e1386c8f3c641fabfaa1182a142f9516a1886404937fd8370" + }, + "checkpoints/checkpoint-224/rng_state.pth": { + "size": 14645, + "sha256": "d5430fd68b32a32b46362508782ccee40b0d4f9e402b5b92f2ec0b887608f4c6" + }, + "checkpoints/checkpoint-224/scheduler.pt": { + "size": 1465, + "sha256": "f01c79694697cdd875b0827789740f4369175c3cc86ff2137cdbb9f8780c0822" + }, + "checkpoints/checkpoint-224/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-224/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-224/tokens_state.json": { + "size": 39, + "sha256": "14c3d2181ef61e0cf34f6fe59f885a012f56477195e0bfcf74f965a091355c99" + }, + "checkpoints/checkpoint-224/trainer_state.json": { + "size": 98228, + "sha256": "8983555ef48957aa23a4d723770ff49bf77c651c2eeaef19db56e010a5d038e4" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-256/adapter_model.safetensors": { + "size": 547777976, + "sha256": "82e30de899bd2e0838c78dbb92e10c762d6a3b99ad47d2af903b54762f9cc7f7" + }, + "checkpoints/checkpoint-256/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-256/optimizer.pt": { + "size": 1048106435, + "sha256": "f0576fcc9b01cf627a5ec161d2c49a7671a4643b01c242b140dec1cd4c134325" + }, + "checkpoints/checkpoint-256/rng_state.pth": { + "size": 14645, + "sha256": "602ecc8706138be0436f114289739d17e95f29812727cd5c1d6cd5c08e5d8786" + }, + "checkpoints/checkpoint-256/scheduler.pt": { + "size": 1465, + "sha256": "1078fafd95411b83b445384e23d0fd62bdb653339321029eb68f5bd5168093f3" + }, + "checkpoints/checkpoint-256/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-256/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-256/tokens_state.json": { + "size": 39, + "sha256": "a1db2747ccfd098f33eee7a195bbd033988a2849ce30491603e0d31cbe794a14" + }, + "checkpoints/checkpoint-256/trainer_state.json": { + "size": 112270, + "sha256": "6382ba31542ad67c894e6feb28e6e22a42b4bc496ec395788ae3f2bf1c54bd4b" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-288/adapter_model.safetensors": { + "size": 547777976, + "sha256": "8018d54994da8531938bc6e3ad65b757ae0133e9558d078a9df3ffcd38a90770" + }, + "checkpoints/checkpoint-288/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-288/optimizer.pt": { + "size": 1048106435, + "sha256": "a675c0e9506d0bcbdd672e493f23320efca843be2204c08bad3906fe3bff8daf" + }, + "checkpoints/checkpoint-288/rng_state.pth": { + "size": 14645, + "sha256": "8af482dc7714cc312f05cbc3db793fa64a2fdbcf28708d6779dc374be7587d64" + }, + "checkpoints/checkpoint-288/scheduler.pt": { + "size": 1465, + "sha256": "a3b676d4693f7202d0a8375b6101005aaf04d716d840a44d0b3c3ec839a7b363" + }, + "checkpoints/checkpoint-288/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-288/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-288/tokens_state.json": { + "size": 39, + "sha256": "f1eed36b4e578b166623815e5accad0aac99aa3b265255a7471d4f0a7c22b297" + }, + "checkpoints/checkpoint-288/trainer_state.json": { + "size": 126367, + "sha256": "03fade8974a90a6a324833d1c68a8002011634ebfdd9232142503f52f1cd57c8" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-32/adapter_model.safetensors": { + "size": 547777976, + "sha256": "ed36ac4554987a7152a477983f7a314c0c9f68f9a8a211f94a73cdf2b705d904" + }, + "checkpoints/checkpoint-32/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-32/optimizer.pt": { + "size": 1048106435, + "sha256": "5d4b6de24e7a07c206a9734f6df37171b7a7981489ce659c718ef9294575b4b8" + }, + "checkpoints/checkpoint-32/rng_state.pth": { + "size": 14645, + "sha256": "9168f0a62aa7cd48f2ccb2c49fb4f47c4f3334e00836cd41f2ecba73cf02d742" + }, + "checkpoints/checkpoint-32/scheduler.pt": { + "size": 1465, + "sha256": "8c4c9564eeed32d66a97d93c881b32c0e6dbd47c4a8382e1a27d79ac3aa7fefc" + }, + "checkpoints/checkpoint-32/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-32/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-32/tokens_state.json": { + "size": 37, + "sha256": "d9025748b8b157d8490d18576a010318e827c122bd7c9d4da1001a4e203d8a9b" + }, + "checkpoints/checkpoint-32/trainer_state.json": { + "size": 14413, + "sha256": "098279468a5e4792a52190ff5aff0c36d3fc4cdc02f7dad0748d9fc5ecbd0257" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-320/adapter_model.safetensors": { + "size": 547777976, + "sha256": "14d37ede82390683030544c61cbe50e026c88a647f59a14da1cc89fcbc5ce882" + }, + "checkpoints/checkpoint-320/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-320/optimizer.pt": { + "size": 1048106435, + "sha256": "f252cbcefa102d5c846b04d29958c1ee5008b71c59145e116f03bdc491ca0312" + }, + "checkpoints/checkpoint-320/rng_state.pth": { + "size": 14645, + "sha256": "0118b4ae79cc2c2b8bd9963e1c8e9e2f135c87084bea8d5d4733a597322339dd" + }, + "checkpoints/checkpoint-320/scheduler.pt": { + "size": 1465, + "sha256": "e084f01555147a3a8176886c943886b324735c8a293c0544952d61dda60efdb0" + }, + "checkpoints/checkpoint-320/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-320/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-320/tokens_state.json": { + "size": 39, + "sha256": "fa2e57cf6dee598825b9bba06159bc70e457d9e8315c9d1028a34b835eec8770" + }, + "checkpoints/checkpoint-320/trainer_state.json": { + "size": 140487, + "sha256": "be060313257db1af22b7936cb09a70fbb5a3da0fac45317665b7de2a7c7af758" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-352/adapter_model.safetensors": { + "size": 547777976, + "sha256": "79438e5bf03bca03b99fdc4416bcd7dff05d12835b9cba1207c438c4fc14ee0f" + }, + "checkpoints/checkpoint-352/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-352/optimizer.pt": { + "size": 1048106435, + "sha256": "a20403b884a94692d1c323bcadd593673a1cc3fbbdac75ad629539d36a018a8a" + }, + "checkpoints/checkpoint-352/rng_state.pth": { + "size": 14645, + "sha256": "cdb0b23f5e396570ae690118f244af63cfde4fa056ae979423e1e2d904a44a02" + }, + "checkpoints/checkpoint-352/scheduler.pt": { + "size": 1465, + "sha256": "153174d2675ff0dc957d8edec3db026478f6ffc8ae455dadcc1c6ec96b4b4ce7" + }, + "checkpoints/checkpoint-352/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-352/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-352/tokens_state.json": { + "size": 40, + "sha256": "b9bc28c20f274eed19841a293251e42a3a606798b019918fd3fd29a96ca3043c" + }, + "checkpoints/checkpoint-352/trainer_state.json": { + "size": 154624, + "sha256": "c3e1a55f1a34c132f45b9fc607f927023e4b7f5f8eba5ec57e699d3414af1f0c" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-384/adapter_model.safetensors": { + "size": 547777976, + "sha256": "686fb1be8b0d19faf30c476f2efbf9cd54b3b548d97d632b918a5d377c09e1af" + }, + "checkpoints/checkpoint-384/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-384/optimizer.pt": { + "size": 1048106435, + "sha256": "19f04c895519cc138e3f57329e87fa6c681b12b8248d1844bffd7413c2d3d840" + }, + "checkpoints/checkpoint-384/rng_state.pth": { + "size": 14645, + "sha256": "d1f8c2ac8d227cb364e0a4240850f857f6a9916ec56e775c26ccc9d6439c6831" + }, + "checkpoints/checkpoint-384/scheduler.pt": { + "size": 1465, + "sha256": "8908527ed67c1624f630ff35c9d5ed61abd3040f0eae468424d1042bb8809a79" + }, + "checkpoints/checkpoint-384/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-384/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-384/tokens_state.json": { + "size": 40, + "sha256": "3075bc59309b4b7b0fefd61404619b2d7b2fe67268f884f30ea4319fd447e21e" + }, + "checkpoints/checkpoint-384/trainer_state.json": { + "size": 168785, + "sha256": "a6acc5ed57dbd0d75fe5616b6c24a604fb9cfb3f48f7dafedec2749b2f730687" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-416/adapter_model.safetensors": { + "size": 547777976, + "sha256": "363d32a4a0f31100662840fbb803e48f26493a0a0d68db2bc8e9ae77eb67ed61" + }, + "checkpoints/checkpoint-416/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-416/optimizer.pt": { + "size": 1048106435, + "sha256": "ab2f0117cf3ff1112d19c4ece704fe24b25018899e51e946efe5a160af5bbc43" + }, + "checkpoints/checkpoint-416/rng_state.pth": { + "size": 14645, + "sha256": "014a4d33d6570bda96dc7af90236c0737fc5b5159f09b97a6588d18406f57688" + }, + "checkpoints/checkpoint-416/scheduler.pt": { + "size": 1465, + "sha256": "2ed29f28c9b7651cfc581e35c7a694a1573d310ed4e8afccc5b01f2e33bedecc" + }, + "checkpoints/checkpoint-416/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-416/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-416/tokens_state.json": { + "size": 40, + "sha256": "033256f14efaa1c491d2eeb0fe5a1f206222c187594393c6791fa23f37aacb5e" + }, + "checkpoints/checkpoint-416/trainer_state.json": { + "size": 182947, + "sha256": "3eb473873b3c44f27b7f2cf0a0535515aed2ed0db743b342631d41e0b1651bd7" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-448/adapter_model.safetensors": { + "size": 547777976, + "sha256": "ddab61d8202f7c1b2b90f50b916d2190d69acc0de38112a102ce9928746082e7" + }, + "checkpoints/checkpoint-448/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-448/optimizer.pt": { + "size": 1048106435, + "sha256": "fad8c2740aa902c6e9c7743ee80e28ea997db6d411814062b88f420ea7e5a07f" + }, + "checkpoints/checkpoint-448/rng_state.pth": { + "size": 14645, + "sha256": "d27d98c90c4450806b1ea70f841a527e6e8f8e0dd5e5e1af25cbd5dd7470342a" + }, + "checkpoints/checkpoint-448/scheduler.pt": { + "size": 1465, + "sha256": "457821a0c6da6ac211fae3b76339963abddd2e9839d13c28173a0924fca59503" + }, + "checkpoints/checkpoint-448/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-448/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-448/tokens_state.json": { + "size": 40, + "sha256": "87ae99c7d9168aee7dd04df6332dedc5884f11b6c9ce320edc6d2b867c048531" + }, + "checkpoints/checkpoint-448/trainer_state.json": { + "size": 197116, + "sha256": "1eff5acfd7d39714a6af0032b8bb327cbb84d9584779b7c3a1d0e54ec54d013d" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-480/adapter_model.safetensors": { + "size": 547777976, + "sha256": "75141177e9bccdf92b1ac95302b87d05dd67560b0535f7f736eead303fe25a9f" + }, + "checkpoints/checkpoint-480/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-480/optimizer.pt": { + "size": 1048106435, + "sha256": "f2b7c832be1bd63d2b26b4e654c51bf44f5012f24a16ebbaacac4825dba48a72" + }, + "checkpoints/checkpoint-480/rng_state.pth": { + "size": 14645, + "sha256": "106674c12c31685485a283c8c514a41fe51a83a990a0c55b807abd9181094cbe" + }, + "checkpoints/checkpoint-480/scheduler.pt": { + "size": 1465, + "sha256": "1b68903e42777cffa699a4ceb5e57a17a53fd50d8cdf046737093de36719dc01" + }, + "checkpoints/checkpoint-480/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-480/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-480/tokens_state.json": { + "size": 40, + "sha256": "6c547605fe97848a0748ea1ac9bef1baa4a78902cfb711aedb151a69c5994f49" + }, + "checkpoints/checkpoint-480/trainer_state.json": { + "size": 211299, + "sha256": "ecb71e590c37b8ae24bf6f819376466d335b61950930f761f3c50b0badcd0b58" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-512/adapter_model.safetensors": { + "size": 547777976, + "sha256": "82e5c558fd3518967c97a463cc23363818705003444ae831b5667d43e013cc03" + }, + "checkpoints/checkpoint-512/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-512/optimizer.pt": { + "size": 1048106435, + "sha256": "906d8543a4a822e6c81e78141cb51fe2165a634f54a2424c75faa38653b40c90" + }, + "checkpoints/checkpoint-512/rng_state.pth": { + "size": 14645, + "sha256": "8b4b9d2fe88b5aebe9f6d0b2a3cb27b73b4b5baa17ae0760b0ccb0011a29ee6d" + }, + "checkpoints/checkpoint-512/scheduler.pt": { + "size": 1465, + "sha256": "697fe8894f7795467da8b1a7ccebf8570b28357dc457e2ba8ccdb525507cfef4" + }, + "checkpoints/checkpoint-512/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-512/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-512/tokens_state.json": { + "size": 40, + "sha256": "c0fcac61392efcd924c1610ad304635be7ae41b33c0ad3526b109f7b1f8f7fe9" + }, + "checkpoints/checkpoint-512/trainer_state.json": { + "size": 225493, + "sha256": "1267db0ee89146bb815e9803774c8f008e757c232cdeb98ee74f3ded2c7a558e" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-64/adapter_model.safetensors": { + "size": 547777976, + "sha256": "430c1e96249f6e4e29049185194cf13579a5a1e6487cb8371f94f231cdb9f181" + }, + "checkpoints/checkpoint-64/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-64/optimizer.pt": { + "size": 1048106435, + "sha256": "e585743619c19db77208c38163c09c890ef8a0d3eccd47e77ca9e6a6e5f9c18f" + }, + "checkpoints/checkpoint-64/rng_state.pth": { + "size": 14645, + "sha256": "e81ea1736533aae9054c1b0b95919202f436afdc19fabdedd95d01577c82dba0" + }, + "checkpoints/checkpoint-64/scheduler.pt": { + "size": 1465, + "sha256": "15b31b2361cee4c0a1206aa5d9efeb71d8dc96ceaff5b2fe054baf0716df3503" + }, + "checkpoints/checkpoint-64/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-64/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-64/tokens_state.json": { + "size": 38, + "sha256": "9c3a6e3ed0949f0de9f9b7c1d666e376e3d3626dfe0a6da2a265ef135b18e10b" + }, + "checkpoints/checkpoint-64/trainer_state.json": { + "size": 28326, + "sha256": "e30f64f9f4f61e414b5b1dcc389edd15d6c6df229f30c0c2fbe8eef68a9035dc" + }, + "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": "1251c6b0693d1cb668f007cc7b0e301a1c2322042b7c1562b8b221a48cf3ddb5" + }, + "checkpoints/checkpoint-96/adapter_model.safetensors": { + "size": 547777976, + "sha256": "168ee78fd358d0507578c9ede2a6393a09754e593fe72279527600dff8a68167" + }, + "checkpoints/checkpoint-96/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-96/optimizer.pt": { + "size": 1048106435, + "sha256": "19c163b6e0ec068b3776ecfff15d083928a6ffad7d6ed366b995cd378ec2bb9e" + }, + "checkpoints/checkpoint-96/rng_state.pth": { + "size": 14645, + "sha256": "42c59541dcd19cf1955908c9bd031702825e10de5b0ff1d1701848480027e815" + }, + "checkpoints/checkpoint-96/scheduler.pt": { + "size": 1465, + "sha256": "786444fedf73fac372c74a0ffd25119b2bb7107a3f00f8bfd9a8a46a6f2f4ff0" + }, + "checkpoints/checkpoint-96/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-96/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-96/tokens_state.json": { + "size": 38, + "sha256": "f276353ce1fa0f6362678bebc659fb55eeca16e731ebeb9a62916aa8efaa3b8b" + }, + "checkpoints/checkpoint-96/trainer_state.json": { + "size": 42250, + "sha256": "46c1862b9a4309c4e59223da5b93b05fede0ae43269e71a12609236affb77121" + }, + "checkpoints/checkpoint-96/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/config.json": { + "size": 3278, + "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a" + }, + "checkpoints/debug.log": { + "size": 269100, + "sha256": "2a563aa608bfe381c560b7e7fbb73677d7f59592afb73a1ba90b04ebd86cf574" + }, + "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_coin_real_1x__agreement.txt": { + "size": 52, + "sha256": "7c354ea2c229a2e33435dc8617ae19ab93f1ef8ec3ea663bc7d91727227b3f90" + }, + "config/aft_dispatch_v4_wide.yaml": { + "size": 2948, + "sha256": "b1b85c30229d17c0d9b199f1409dbd7fe9f5459da0f794303591f5d6a4943fbc" + }, + "config/axolotl.yaml": { + "size": 1211, + "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e" + }, + "health/training_started.json": { + "size": 137, + "sha256": "a61a63b0cd3e21aa09ce4870d83278752242f6a12692273628067ecf3c9dca17" + }, + "run.json": { + "size": 407, + "sha256": "ba87850992cb9947acffc29fa80fdf2097d77ad225a6f88ede61aafbe6e98ef5" + }, + "train.log": { + "size": 276492, + "sha256": "85098917d00df3a281a4a67f95305f448dee02b660b71a07a8acb96739efecdc" + }, + "trainer_state.final.json": { + "size": 225493, + "sha256": "1267db0ee89146bb815e9803774c8f008e757c232cdeb98ee74f3ded2c7a558e" + }, + "training_examples.jsonl": { + "size": 3159083, + "sha256": "cf7a6dbf26a4dc12581e11f94a2229dfb1c1d7a487dc666fa42b49977bd2b5db" + }, + "training_provenance.json": { + "size": 4090, + "sha256": "916bbf8d14dc79f693d74f3a89c7e869573f45ffbca7aa099cf6b69c3bec7f1d" + }, + "training_trace.jsonl": { + "size": 182225, + "sha256": "9b8095748b11eebe446033d837bbc8ca237031cc4a1bff709ad8d35a0bbc8ea0" + } + } +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/TRAINED.json b/aft_wave_v2/coin_real_1x__agreement/training/TRAINED.json new file mode 100644 index 0000000000000000000000000000000000000000..da416a61942d7d1190bad969e02c2d1136a8eeb6 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/TRAINED.json @@ -0,0 +1,55 @@ +{ + "version": "dispatch_wave_v2", + "arm": "coin_real_1x__agreement", + "parameterization": "lora", + "parent_repo": "arcadia-impact/scimt-dispatch-models", + "parent_prefix": "sft/coin/checkpoint-48", + "dataset_sha256": "8f28a074352168b89e47c6555e9c2036f2c6e79903bbd588dbb7972fd57b5e2b", + "training_rows": 8192, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "minutes": 58.42, + "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/coin_real_1x__agreement/training/axolotl.yaml b/aft_wave_v2/coin_real_1x__agreement/training/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8e9757f2139725f1db19ed2fb1e692ba1d928e0 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/axolotl.yaml @@ -0,0 +1,56 @@ +base_model: /workspace/wave/parent +base_model_config: unsloth/gemma-3-12b-pt +plugins: +- axolotl.integrations.liger.LigerPlugin +liger_fused_linear_cross_entropy: true +liger_rope: true +liger_rms_norm: true +liger_glu_activation: true +datasets: +- path: /workspace/wave/data/datasets/aft_agreement.jsonl + type: chat_template + field_messages: messages +eot_tokens: +- +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/coin_real_1x__agreement/training/checkpoint.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoint.json new file mode 100644 index 0000000000000000000000000000000000000000..bd6ca7c80d3807fe4675bdc282acd95f25090ae7 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoint.json @@ -0,0 +1,74 @@ +{ + "experiment": "scimt-train:coin_real_1x__agreement", + "spec": null, + "kind": null, + "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.", + "train": { + "data": "/workspace/wave/data/datasets/aft_agreement.jsonl", + "dataset_meta": { + "adhoc": true + }, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "load_checkpoint_path": "/workspace/wave/parent", + "grpo": null, + "lora": { + "r": 32, + "alpha": 64, + "dropout": 0.05, + "target_linear": false, + "target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "initial_adapter_path": null + } + }, + "run_name": "coin_real_1x__agreement", + "pointer_file": "/workspace/wave/training/ckpt_coin_real_1x__agreement.txt", + "backend": "axolotl", + "sampler": "/workspace/wave/training/checkpoints/checkpoint-512", + "state": "/workspace/wave/training/checkpoints/checkpoint-512", + "model": "gemma3_12b_it", + "meta": { + "experiment": "scimt-train:coin_real_1x__agreement", + "spec": null, + "kind": null, + "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.", + "train": { + "data": "/workspace/wave/data/datasets/aft_agreement.jsonl", + "dataset_meta": { + "adhoc": true + }, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "load_checkpoint_path": "/workspace/wave/parent", + "grpo": null, + "lora": { + "r": 32, + "alpha": 64, + "dropout": 0.05, + "target_linear": false, + "target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "initial_adapter_path": null + } + }, + "run_name": "coin_real_1x__agreement", + "pointer_file": "/workspace/wave/training/ckpt_coin_real_1x__agreement.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/coin_real_1x__agreement/training/checkpoints.jsonl b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..0a5fa00d7287f6920786ad43da67f36b8568e07a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/README.md new file mode 100644 index 0000000000000000000000000000000000000000..ca90b96febc25002c5ca143b930366070605caa3 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/README.md @@ -0,0 +1,128 @@ +--- +library_name: peft +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +datasets: +- /workspace/wave/data/datasets/aft_agreement.jsonl +pipeline_tag: text-generation +base_model: /workspace/wave/parent +model-index: +- name: workspace/wave/training/checkpoints + results: [] +--- + + + +[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_agreement.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_agreement.jsonl dataset. + +## Model description + +More information needed + +## Intended uses & limitations + +More information needed + +## Training and evaluation data + +More information needed + +## Training procedure + +### Training hyperparameters + +The following hyperparameters were used during training: +- learning_rate: 0.0001 +- train_batch_size: 16 +- eval_batch_size: 16 +- seed: 42 +- gradient_accumulation_steps: 2 +- total_train_batch_size: 32 +- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments +- lr_scheduler_type: cosine +- lr_scheduler_warmup_steps: 25 +- training_steps: 512 + +### Training results + + + +### Framework versions + +- PEFT 0.19.1 +- Transformers 5.9.0 +- Pytorch 2.12.1+cu126 +- Datasets 4.8.5 +- Tokenizers 0.22.2 \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..32a98d0c896ae739414f6dad378cf34abb424d19 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82e5c558fd3518967c97a463cc23363818705003444ae831b5667d43e013cc03 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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/coin_real_1x__agreement/training/checkpoints/checkpoint-128/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## 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/coin_real_1x__agreement/training/checkpoints/checkpoint-128/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..cc3e732d462681a1e3d2d619c243dcda4cf21a08 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d6f7dc703af7a80756c2db7f916e936ba8f3fb1888e9774543777019deda638 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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/coin_real_1x__agreement/training/checkpoints/checkpoint-128/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c23f9e2e67d7d04ce35850753c05b7169138ee7b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:368b3dac3ce73c64c978a430804a3a64d276128c43f4b8d6f31d0f5379cb36e8 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..d1b1a7f801175d6cdfcf91c20e430f7f95dfbcc2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9983d79410c5ae3ad4dd6b33d87b229e318829b73aa2e270784b5be7d242f714 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc9eb8d587c7c4d2a9e80caed8c0953081dce4ba --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6dc656b5184553709bb4f936a6e1af19bc4036e0 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/tokens_state.json @@ -0,0 +1 @@ +{"total": 3869824, "trainable": 58534} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..1f1dff4f81bd999c0bbb24ccf93549bdbdeff983 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/trainer_state.json @@ -0,0 +1,1826 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.5, + "eval_steps": 500, + "global_step": 128, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + } + ], + "logging_steps": 1, + "max_steps": 512, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 32, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 2.6266756914078106e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-128/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## 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/coin_real_1x__agreement/training/checkpoints/checkpoint-160/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f61d9adc7c39b908c8207116958d8bc0921bd026 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fab5da99ffb1e531fe46e097946bd2e7cfe29250a2377f2e028f7034be1388c0 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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/coin_real_1x__agreement/training/checkpoints/checkpoint-160/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..47ea8ce93b391ed36179967d6bff7560242c44f9 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62dfc9938a2107ed60a481bb92f478c5b103d1cb7499a5be2252d6378799c68c +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..a6d6743cfc4c8048028999ac0a30aaa5d2704164 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbe19eb8e9b9385d32e98b493727887e743dd30947d7e742ac3b5158b9819536 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..d47c908eab228a9c923b0bfb0a7fb1a6dfa773e6 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..f11ef3bc52964a4473f975e5cfbdff990f797946 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/tokens_state.json @@ -0,0 +1 @@ +{"total": 4840112, "trainable": 73148} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3fd307d16b394b2d05f0fdae8679bc16ef932996 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/trainer_state.json @@ -0,0 +1,2274 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.625, + "eval_steps": 500, + "global_step": 160, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + } + ], + "logging_steps": 1, + "max_steps": 512, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 32, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 3.285266858154593e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-160/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## 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/coin_real_1x__agreement/training/checkpoints/checkpoint-192/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7e945fe3739522fd68c1e83c36e6adab6a166692 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a56c1fbd3eb063bfa0ce7a15a20577569c29fcec60eb4c20da60cbdab95b19dc +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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/coin_real_1x__agreement/training/checkpoints/checkpoint-192/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..cd63e84f56ec35f97e88e8df12c89dbf6ddff88a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c2141d2e5550443cbf4a5cdf9f0cfb92cc5cd62fd8cf95f4c09d3fc8a321fdd +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..6a619c1a5d95edb5ac07c77f089d1d3ef6cb25f2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d896644f2e98c81e623938a072f96d9cb9eb180436ba03e68153edafca3bfcb +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..68fafd70d4f03fc26bd72aefbdcd22105c983415 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..620715ae21d29fb1fcb2456e445b5b58f0a3d82a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/tokens_state.json @@ -0,0 +1 @@ +{"total": 5803136, "trainable": 87747} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..16868cbed7e3bb98ff5d58cd380d1ef3a9d4c8b4 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/trainer_state.json @@ -0,0 +1,2722 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.75, + "eval_steps": 500, + "global_step": 192, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + } + ], + "logging_steps": 1, + "max_steps": 512, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 32, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 3.938927523611812e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-192/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## 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/coin_real_1x__agreement/training/checkpoints/checkpoint-224/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6827a2ce9d014522222c9b60903633cb0a600d97 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26345eeb816e18811525163b2b729f6c6d1d7c1cd6d604d373815998d830ab55 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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/coin_real_1x__agreement/training/checkpoints/checkpoint-224/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..dc941d906ad08f819e2f9c022b2993a242ebafcb --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8970aa983a72b53e1386c8f3c641fabfaa1182a142f9516a1886404937fd8370 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..1f31e67db8f0d39a135e0639466ef963bbf0e89a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5430fd68b32a32b46362508782ccee40b0d4f9e402b5b92f2ec0b887608f4c6 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..734e9a25165547994d3add43bbfb2dd65b7e559d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..483ca1252563c6374c89ccfdd65ff7bd08c5b70c --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/tokens_state.json @@ -0,0 +1 @@ +{"total": 6774416, "trainable": 102411} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..0d8eb1a8b6be4e38b494f1ac00ee81aea91bca51 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + } + ], + "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": 4.598192018728535e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-224/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-256/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-256/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e6798d26bec68634b210c83e1e735aff97038aec --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82e30de899bd2e0838c78dbb92e10c762d6a3b99ad47d2af903b54762f9cc7f7 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-256/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..85393afd167c4ec23439f87fd60c3f57a6ed4c06 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0576fcc9b01cf627a5ec161d2c49a7671a4643b01c242b140dec1cd4c134325 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..636b6a233211b6590d46b3e4472b2c251d7e3ffc --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:602ecc8706138be0436f114289739d17e95f29812727cd5c1d6cd5c08e5d8786 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..0d402b407a8fd89f8bcb5de6f509c9979553d18a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..45bb3fa28e090bed085f25572eb045bc065f08b2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/tokens_state.json @@ -0,0 +1 @@ +{"total": 7745872, "trainable": 117030} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c99b75a1227f72d779292ca6e6cc84b50284b43d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + } + ], + "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": 5.257575975330248e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-256/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-288/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-288/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..efd9a1947844c45642fbde57fa26e113f76b0931 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8018d54994da8531938bc6e3ad65b757ae0133e9558d078a9df3ffcd38a90770 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-288/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..8a770f9917d2439fe982cf3dff0d3c84cb397d57 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a675c0e9506d0bcbdd672e493f23320efca843be2204c08bad3906fe3bff8daf +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..ec4d43c2bb22b990826392ae2b797407fc091593 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8af482dc7714cc312f05cbc3db793fa64a2fdbcf28708d6779dc374be7587d64 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f00d58c65c626a11e09272704f772fedb532412 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fb5e8bfc6e6bef0e3658333623e6da3328ca8413 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/tokens_state.json @@ -0,0 +1 @@ +{"total": 8713504, "trainable": 131763} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..0a56945c1c572411d42acacb3cd6507a908b3ff6 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + } + ], + "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": 5.91436435966719e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-288/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-32/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-32/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..dcc49367f83f82013232351b97d661dcb5e267d1 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ed36ac4554987a7152a477983f7a314c0c9f68f9a8a211f94a73cdf2b705d904 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-32/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c24bb315129902ce1a90e5b298cad15c54acddba --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5d4b6de24e7a07c206a9734f6df37171b7a7981489ce659c718ef9294575b4b8 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..90709c63860f53d1463dc2e454795eef420bf323 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9168f0a62aa7cd48f2ccb2c49fb4f47c4f3334e00836cd41f2ecba73cf02d742 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..02f9ed8e0defc2ebe0d43c0d14abc27c32e2b2cf --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..bc9cd4715a9ff75ffee41970831490722066f050 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/tokens_state.json @@ -0,0 +1 @@ +{"total": 964832, "trainable": 14565} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2402d8832ac22ef8fc2941d1d0eaba636bd6b84f --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + } + ], + "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": 6.548878607121101e+16, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-32/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-320/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-320/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..70ff89c29d4d2230b955f6342e1a2282d83a9bb4 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14d37ede82390683030544c61cbe50e026c88a647f59a14da1cc89fcbc5ce882 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-320/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..126b71d613549b03cefc281f142d21cb64bff883 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f252cbcefa102d5c846b04d29958c1ee5008b71c59145e116f03bdc491ca0312 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..718496431dd7f3ae281f4484cb2b0ec3cb7faf22 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0118b4ae79cc2c2b8bd9963e1c8e9e2f135c87084bea8d5d4733a597322339dd +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf4669ac28031073c15cff33e6a2bd7daa0e7337 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..abc36ad05608effc5334e440987c46913cd84289 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/tokens_state.json @@ -0,0 +1 @@ +{"total": 9682224, "trainable": 146314} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..5e7ce44157285d998447441789714cf2e4a53c52 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.004979890305548906, + "learning_rate": 4.9354244499126866e-05, + "loss": 3.968320743297227e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0025452408008277416, + "learning_rate": 4.90663667927174e-05, + "loss": 3.7232457543723285e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00004, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0009856978431344032, + "learning_rate": 4.877873600900581e-05, + "loss": 2.570721335359849e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0006475941045209765, + "learning_rate": 4.849136411748306e-05, + "loss": 1.7373771697748452e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.2577264606952667, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0018721828237175941, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00187, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.93, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.09887456148862839, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0007890360429883003, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00079, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.005742037668824196, + "learning_rate": 4.7630921326368736e-05, + "loss": 5.973849329166114e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.0062770843505859375, + "learning_rate": 4.7344704475577916e-05, + "loss": 7.13710323907435e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.009947159327566624, + "learning_rate": 4.705880619288153e-05, + "loss": 0.00010848060628632084, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.41, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.006360637955367565, + "learning_rate": 4.677323837567412e-05, + "loss": 5.382625022321008e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.030510911718010902, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00030309066642075777, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0003, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.004953032359480858, + "learning_rate": 4.620314165804964e-05, + "loss": 5.149709977558814e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.003854556707665324, + "learning_rate": 4.591863648168407e-05, + "loss": 4.885891394224018e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.11261430382728577, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0007564757252112031, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00076, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.03299031779170036, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00041300299926660955, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00041, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.0065305293537676334, + "learning_rate": 4.506743570677743e-05, + "loss": 7.41428229957819e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.005171677563339472, + "learning_rate": 4.478451305763618e-05, + "loss": 6.253476749407127e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.006580202840268612, + "learning_rate": 4.450201551660454e-05, + "loss": 5.272469570627436e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00005, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.31961339712142944, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.003208739683032036, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00321, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.0023711167741566896, + "learning_rate": 4.393834276419352e-05, + "loss": 3.434348400332965e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.019930923357605934, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00012936958228237927, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00013, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0023675146512687206, + "learning_rate": 4.337651127546448e-05, + "loss": 2.4802844563964754e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0036730507854372263, + "learning_rate": 4.3096315242201736e-05, + "loss": 3.144190486636944e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0003704871633090079, + "learning_rate": 4.2816614569850635e-05, + "loss": 8.931780030252412e-06, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.43, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.05240330100059509, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00026115920627489686, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00026, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.0012482519960030913, + "learning_rate": 4.225874584473174e-05, + "loss": 2.5392351744812913e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.050520945340394974, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0006884933682158589, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00069, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.002656160155311227, + "learning_rate": 4.170299795992081e-05, + "loss": 2.599508843559306e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.19733326137065887, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0016555915353819728, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00166, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.008795367553830147, + "learning_rate": 4.114946342220728e-05, + "loss": 6.713026959914714e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.00049841299187392, + "learning_rate": 4.087355496656321e-05, + "loss": 1.1344342965458054e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.23748824000358582, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0030270391143858433, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00303, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.19, + "tokens/trainable": 146314 + } + ], + "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": 6.571891233184067e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-320/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-352/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-352/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bd0620e1b5a8e3792e4a21ff3eb9bff5191ef9bb --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:79438e5bf03bca03b99fdc4416bcd7dff05d12835b9cba1207c438c4fc14ee0f +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-352/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..0847ea62e3500e88ca1016fa3be503d86fe36585 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a20403b884a94692d1c323bcadd593673a1cc3fbbdac75ad629539d36a018a8a +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..45ed2fad2b605a1fef83ba070f709a6a614266a5 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cdb0b23f5e396570ae690118f244af63cfde4fa056ae979423e1e2d904a44a02 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..1d2b8d34c008174b230d6dbbee4469d934686b9f --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c0fba86cc090d2b6f9d01156de76a604b849b27e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/tokens_state.json @@ -0,0 +1 @@ +{"total": 10653632, "trainable": 160952} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c79b54bbabcf0f423ab9d6f0304299d5cdcdc673 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.004979890305548906, + "learning_rate": 4.9354244499126866e-05, + "loss": 3.968320743297227e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0025452408008277416, + "learning_rate": 4.90663667927174e-05, + "loss": 3.7232457543723285e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00004, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0009856978431344032, + "learning_rate": 4.877873600900581e-05, + "loss": 2.570721335359849e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0006475941045209765, + "learning_rate": 4.849136411748306e-05, + "loss": 1.7373771697748452e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.2577264606952667, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0018721828237175941, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00187, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.93, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.09887456148862839, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0007890360429883003, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00079, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.005742037668824196, + "learning_rate": 4.7630921326368736e-05, + "loss": 5.973849329166114e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.0062770843505859375, + "learning_rate": 4.7344704475577916e-05, + "loss": 7.13710323907435e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.009947159327566624, + "learning_rate": 4.705880619288153e-05, + "loss": 0.00010848060628632084, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.41, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.006360637955367565, + "learning_rate": 4.677323837567412e-05, + "loss": 5.382625022321008e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.030510911718010902, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00030309066642075777, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0003, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.004953032359480858, + "learning_rate": 4.620314165804964e-05, + "loss": 5.149709977558814e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.003854556707665324, + "learning_rate": 4.591863648168407e-05, + "loss": 4.885891394224018e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.11261430382728577, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0007564757252112031, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00076, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.03299031779170036, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00041300299926660955, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00041, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.0065305293537676334, + "learning_rate": 4.506743570677743e-05, + "loss": 7.41428229957819e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.005171677563339472, + "learning_rate": 4.478451305763618e-05, + "loss": 6.253476749407127e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.006580202840268612, + "learning_rate": 4.450201551660454e-05, + "loss": 5.272469570627436e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00005, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.31961339712142944, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.003208739683032036, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00321, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.0023711167741566896, + "learning_rate": 4.393834276419352e-05, + "loss": 3.434348400332965e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.019930923357605934, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00012936958228237927, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00013, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0023675146512687206, + "learning_rate": 4.337651127546448e-05, + "loss": 2.4802844563964754e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0036730507854372263, + "learning_rate": 4.3096315242201736e-05, + "loss": 3.144190486636944e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0003704871633090079, + "learning_rate": 4.2816614569850635e-05, + "loss": 8.931780030252412e-06, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.43, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.05240330100059509, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00026115920627489686, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00026, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.0012482519960030913, + "learning_rate": 4.225874584473174e-05, + "loss": 2.5392351744812913e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.050520945340394974, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0006884933682158589, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00069, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.002656160155311227, + "learning_rate": 4.170299795992081e-05, + "loss": 2.599508843559306e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.19733326137065887, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0016555915353819728, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00166, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.008795367553830147, + "learning_rate": 4.114946342220728e-05, + "loss": 6.713026959914714e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.00049841299187392, + "learning_rate": 4.087355496656321e-05, + "loss": 1.1344342965458054e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.23748824000358582, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0030270391143858433, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00303, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.19, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0024427378084510565, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.495265627861954e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00003, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.01809818297624588, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00013907425454817712, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0010443986393511295, + "learning_rate": 3.977591418134619e-05, + "loss": 1.951586818904616e-05, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.009523316286504269, + "learning_rate": 3.95030593412612e-05, + "loss": 6.634376040892676e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00007, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.049099843949079514, + "learning_rate": 3.923084939213296e-05, + "loss": 0.00039430244942195714, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 32.11, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.01699669286608696, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00013168745499569923, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00013, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.07120174169540405, + "learning_rate": 3.868840945050728e-05, + "loss": 0.000350497430190444, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00035, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.001589596620760858, + "learning_rate": 3.841820203114995e-05, + "loss": 2.160290750907734e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.014652811922132969, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00011893574264831841, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.032810863107442856, + "learning_rate": 3.787986851704667e-05, + "loss": 0.00019718095427379012, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.02765781246125698, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00016869997489266098, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00017, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.01766805537045002, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00014554498193319887, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00015, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.05623512342572212, + "learning_rate": 3.707773935267552e-05, + "loss": 0.00024310205481015146, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00024, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.0034802353475242853, + "learning_rate": 3.68118397962661e-05, + "loss": 3.652479063021019e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 31.03, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.0005486137815751135, + "learning_rate": 3.654669712344384e-05, + "loss": 1.2760298886860255e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.053988587111234665, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004493523156270385, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00045, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.61, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.023256463930010796, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0001921930379467085, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00019, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.87, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.005770122166723013, + "learning_rate": 3.575592058295017e-05, + "loss": 5.241332473815419e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0003110536199528724, + "learning_rate": 3.549391545931585e-05, + "loss": 5.897365099372109e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.008365475572645664, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.736510451650247e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00008, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.003012418746948242, + "learning_rate": 3.49723512647657e-05, + "loss": 3.9521881262771785e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.00036992691457271576, + "learning_rate": 3.471281389826491e-05, + "loss": 9.513973964203615e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.77, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.004851094912737608, + "learning_rate": 3.4454120764386764e-05, + "loss": 3.941698741982691e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.08602030575275421, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.0006739848759025335, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00067, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.007851852104067802, + "learning_rate": 3.3939310220027456e-05, + "loss": 7.82167335273698e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00008, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.34, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.000522423128131777, + "learning_rate": 3.3683214232914404e-05, + "loss": 7.171449851739453e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00001, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0008156000403687358, + "learning_rate": 3.342800532426873e-05, + "loss": 1.2959781088284217e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.012755601666867733, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00010119444777956232, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0001, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.0019102037185803056, + "learning_rate": 3.292029118616024e-05, + "loss": 2.8528616894618608e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.07546474039554596, + "learning_rate": 3.266780708475511e-05, + "loss": 0.0005027748411521316, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0005, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.014789975248277187, + "learning_rate": 3.241625231705354e-05, + "loss": 0.00011153960804222152, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.001108153141103685, + "learning_rate": 3.216563735127618e-05, + "loss": 1.160802275990136e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 160952 + } + ], + "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": 7.231242609380782e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-352/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-384/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-384/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..91dd9751d86645c10051759f0a5fa8299099c9b4 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:686fb1be8b0d19faf30c476f2efbf9cd54b3b548d97d632b918a5d377c09e1af +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-384/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..4ce5b20cc68d1f4ac34f197bd9387793543e1fe1 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19f04c895519cc138e3f57329e87fa6c681b12b8248d1844bffd7413c2d3d840 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..d98bb2e9db712c9e711f32bc1e3142fe2e3e490b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1f8c2ac8d227cb364e0a4240850f857f6a9916ec56e775c26ccc9d6439c6831 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c676299e10f0344839a2b08f1cc1e004410cac0 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2d442c4354d176e51ea6a8521d49ac7a4d4dfa22 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/tokens_state.json @@ -0,0 +1 @@ +{"total": 11622000, "trainable": 175586} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..51f2f3a7640541215dbc9a3ff188f1d0389988ab --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.004979890305548906, + "learning_rate": 4.9354244499126866e-05, + "loss": 3.968320743297227e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0025452408008277416, + "learning_rate": 4.90663667927174e-05, + "loss": 3.7232457543723285e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00004, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0009856978431344032, + "learning_rate": 4.877873600900581e-05, + "loss": 2.570721335359849e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0006475941045209765, + "learning_rate": 4.849136411748306e-05, + "loss": 1.7373771697748452e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.2577264606952667, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0018721828237175941, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00187, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.93, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.09887456148862839, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0007890360429883003, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00079, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.005742037668824196, + "learning_rate": 4.7630921326368736e-05, + "loss": 5.973849329166114e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.0062770843505859375, + "learning_rate": 4.7344704475577916e-05, + "loss": 7.13710323907435e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.009947159327566624, + "learning_rate": 4.705880619288153e-05, + "loss": 0.00010848060628632084, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.41, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.006360637955367565, + "learning_rate": 4.677323837567412e-05, + "loss": 5.382625022321008e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.030510911718010902, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00030309066642075777, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0003, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.004953032359480858, + "learning_rate": 4.620314165804964e-05, + "loss": 5.149709977558814e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.003854556707665324, + "learning_rate": 4.591863648168407e-05, + "loss": 4.885891394224018e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.11261430382728577, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0007564757252112031, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00076, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.03299031779170036, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00041300299926660955, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00041, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.0065305293537676334, + "learning_rate": 4.506743570677743e-05, + "loss": 7.41428229957819e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.005171677563339472, + "learning_rate": 4.478451305763618e-05, + "loss": 6.253476749407127e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.006580202840268612, + "learning_rate": 4.450201551660454e-05, + "loss": 5.272469570627436e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00005, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.31961339712142944, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.003208739683032036, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00321, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.0023711167741566896, + "learning_rate": 4.393834276419352e-05, + "loss": 3.434348400332965e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.019930923357605934, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00012936958228237927, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00013, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0023675146512687206, + "learning_rate": 4.337651127546448e-05, + "loss": 2.4802844563964754e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0036730507854372263, + "learning_rate": 4.3096315242201736e-05, + "loss": 3.144190486636944e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0003704871633090079, + "learning_rate": 4.2816614569850635e-05, + "loss": 8.931780030252412e-06, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.43, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.05240330100059509, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00026115920627489686, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00026, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.0012482519960030913, + "learning_rate": 4.225874584473174e-05, + "loss": 2.5392351744812913e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.050520945340394974, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0006884933682158589, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00069, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.002656160155311227, + "learning_rate": 4.170299795992081e-05, + "loss": 2.599508843559306e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.19733326137065887, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0016555915353819728, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00166, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.008795367553830147, + "learning_rate": 4.114946342220728e-05, + "loss": 6.713026959914714e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.00049841299187392, + "learning_rate": 4.087355496656321e-05, + "loss": 1.1344342965458054e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.23748824000358582, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0030270391143858433, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00303, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.19, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0024427378084510565, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.495265627861954e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00003, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.01809818297624588, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00013907425454817712, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0010443986393511295, + "learning_rate": 3.977591418134619e-05, + "loss": 1.951586818904616e-05, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.009523316286504269, + "learning_rate": 3.95030593412612e-05, + "loss": 6.634376040892676e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00007, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.049099843949079514, + "learning_rate": 3.923084939213296e-05, + "loss": 0.00039430244942195714, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 32.11, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.01699669286608696, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00013168745499569923, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00013, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.07120174169540405, + "learning_rate": 3.868840945050728e-05, + "loss": 0.000350497430190444, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00035, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.001589596620760858, + "learning_rate": 3.841820203114995e-05, + "loss": 2.160290750907734e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.014652811922132969, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00011893574264831841, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.032810863107442856, + "learning_rate": 3.787986851704667e-05, + "loss": 0.00019718095427379012, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.02765781246125698, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00016869997489266098, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00017, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.01766805537045002, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00014554498193319887, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00015, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.05623512342572212, + "learning_rate": 3.707773935267552e-05, + "loss": 0.00024310205481015146, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00024, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.0034802353475242853, + "learning_rate": 3.68118397962661e-05, + "loss": 3.652479063021019e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 31.03, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.0005486137815751135, + "learning_rate": 3.654669712344384e-05, + "loss": 1.2760298886860255e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.053988587111234665, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004493523156270385, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00045, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.61, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.023256463930010796, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0001921930379467085, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00019, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.87, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.005770122166723013, + "learning_rate": 3.575592058295017e-05, + "loss": 5.241332473815419e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0003110536199528724, + "learning_rate": 3.549391545931585e-05, + "loss": 5.897365099372109e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.008365475572645664, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.736510451650247e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00008, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.003012418746948242, + "learning_rate": 3.49723512647657e-05, + "loss": 3.9521881262771785e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.00036992691457271576, + "learning_rate": 3.471281389826491e-05, + "loss": 9.513973964203615e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.77, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.004851094912737608, + "learning_rate": 3.4454120764386764e-05, + "loss": 3.941698741982691e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.08602030575275421, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.0006739848759025335, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00067, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.007851852104067802, + "learning_rate": 3.3939310220027456e-05, + "loss": 7.82167335273698e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00008, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.34, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.000522423128131777, + "learning_rate": 3.3683214232914404e-05, + "loss": 7.171449851739453e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00001, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0008156000403687358, + "learning_rate": 3.342800532426873e-05, + "loss": 1.2959781088284217e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.012755601666867733, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00010119444777956232, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0001, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.0019102037185803056, + "learning_rate": 3.292029118616024e-05, + "loss": 2.8528616894618608e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.07546474039554596, + "learning_rate": 3.266780708475511e-05, + "loss": 0.0005027748411521316, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0005, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.014789975248277187, + "learning_rate": 3.241625231705354e-05, + "loss": 0.00011153960804222152, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.001108153141103685, + "learning_rate": 3.216563735127618e-05, + "loss": 1.160802275990136e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.061208564788103104, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0005815322510898113, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00058, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.0031761936843395233, + "learning_rate": 3.166726850239794e-05, + "loss": 2.453591878293082e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0003757224476430565, + "learning_rate": 3.141953535845912e-05, + "loss": 6.875088274682639e-06, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0009734284249134362, + "learning_rate": 3.11727834939056e-05, + "loss": 9.81441462499788e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.0010252115316689014, + "learning_rate": 3.092702317708967e-05, + "loss": 8.114137017400935e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.0002301523054484278, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.338565642887261e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00001, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.0006164724472910166, + "learning_rate": 3.0438518053342407e-05, + "loss": 8.090713890851475e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0006751597975380719, + "learning_rate": 3.0195793575103266e-05, + "loss": 1.1418029316700995e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.00013061391655355692, + "learning_rate": 2.9954101301140146e-05, + "loss": 3.586079174056067e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.001630443730391562, + "learning_rate": 2.9713451289255123e-05, + "loss": 1.7084981664083898e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00002, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0003005889884661883, + "learning_rate": 2.9473853553877484e-05, + "loss": 6.793928150727879e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.005686561111360788, + "learning_rate": 2.9235318065647e-05, + "loss": 5.100644557387568e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.43, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.0014047786826267838, + "learning_rate": 2.8997854750998964e-05, + "loss": 9.773429155757185e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 9.058098657988012e-05, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.382128286626539e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.0, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.0009709852165542543, + "learning_rate": 2.8526184124692883e-05, + "loss": 7.577294582006289e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.89, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.0034374946262687445, + "learning_rate": 2.829199644117484e-05, + "loss": 3.723455301951617e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.0033002276904881, + "learning_rate": 2.8058920186702553e-05, + "loss": 2.274679354741238e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.0006988391978666186, + "learning_rate": 2.782696506053033e-05, + "loss": 5.0486655709391925e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.0052696517668664455, + "learning_rate": 2.7596140715257824e-05, + "loss": 2.7391106414142996e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00003, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.0015628528781235218, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.4505098079098389e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.0021482163574546576, + "learning_rate": 2.7137922742128486e-05, + "loss": 1.399279062752612e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.001900387927889824, + "learning_rate": 2.691054818259188e-05, + "loss": 1.4173308954923414e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 35.04, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.003817511023953557, + "learning_rate": 2.6684342539801933e-05, + "loss": 2.5515182642266154e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00003, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.3609258234500885, + "learning_rate": 2.645931522709877e-05, + "loss": 0.006828296929597855, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00685, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 41.02, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.0038110739551484585, + "learning_rate": 2.6235475608787365e-05, + "loss": 3.2214695238508284e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0038299988955259323, + "learning_rate": 2.6012832999747916e-05, + "loss": 1.8610287952469662e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00002, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.24727243185043335, + "learning_rate": 2.579139666504821e-05, + "loss": 0.029269637539982796, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0297, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.08204144984483719, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0006305916467681527, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00063, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.016210485249757767, + "learning_rate": 2.5352179627565532e-05, + "loss": 9.964105993276462e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0001, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.99, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.00014741058112122118, + "learning_rate": 2.5134417202396277e-05, + "loss": 4.461833668756299e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.00043298961827531457, + "learning_rate": 2.491789760603361e-05, + "loss": 1.0579917216091417e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.0009353766799904406, + "learning_rate": 2.4702629848741764e-05, + "loss": 1.3036698874202557e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 175586 + } + ], + "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": 7.88853055992768e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-384/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-416/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-416/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2eb2ac0b0825f9ef12517a4bc153fa8176fa435c --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:363d32a4a0f31100662840fbb803e48f26493a0a0d68db2bc8e9ae77eb67ed61 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-416/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..e554b1dc21f3e6424cf34de814726d80e58c0dc0 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ab2f0117cf3ff1112d19c4ece704fe24b25018899e51e946efe5a160af5bbc43 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..378ac734ffa403364eeb8e241ac5d35631fe136e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:014a4d33d6570bda96dc7af90236c0737fc5b5159f09b97a6588d18406f57688 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..20a981d8c8f57b8d2e18429ceb299414229ad6ac --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..ca38c2c0f858d125af38a696e17cd1c5dc72b78e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/tokens_state.json @@ -0,0 +1 @@ +{"total": 12590736, "trainable": 190306} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..0fcf65dcf24510a3f637f1996da4326c17dd6441 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.004979890305548906, + "learning_rate": 4.9354244499126866e-05, + "loss": 3.968320743297227e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0025452408008277416, + "learning_rate": 4.90663667927174e-05, + "loss": 3.7232457543723285e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00004, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0009856978431344032, + "learning_rate": 4.877873600900581e-05, + "loss": 2.570721335359849e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0006475941045209765, + "learning_rate": 4.849136411748306e-05, + "loss": 1.7373771697748452e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.2577264606952667, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0018721828237175941, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00187, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.93, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.09887456148862839, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0007890360429883003, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00079, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.005742037668824196, + "learning_rate": 4.7630921326368736e-05, + "loss": 5.973849329166114e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.0062770843505859375, + "learning_rate": 4.7344704475577916e-05, + "loss": 7.13710323907435e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.009947159327566624, + "learning_rate": 4.705880619288153e-05, + "loss": 0.00010848060628632084, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.41, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.006360637955367565, + "learning_rate": 4.677323837567412e-05, + "loss": 5.382625022321008e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.030510911718010902, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00030309066642075777, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0003, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.004953032359480858, + "learning_rate": 4.620314165804964e-05, + "loss": 5.149709977558814e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.003854556707665324, + "learning_rate": 4.591863648168407e-05, + "loss": 4.885891394224018e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.11261430382728577, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0007564757252112031, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00076, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.03299031779170036, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00041300299926660955, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00041, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.0065305293537676334, + "learning_rate": 4.506743570677743e-05, + "loss": 7.41428229957819e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.005171677563339472, + "learning_rate": 4.478451305763618e-05, + "loss": 6.253476749407127e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.006580202840268612, + "learning_rate": 4.450201551660454e-05, + "loss": 5.272469570627436e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00005, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.31961339712142944, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.003208739683032036, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00321, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.0023711167741566896, + "learning_rate": 4.393834276419352e-05, + "loss": 3.434348400332965e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.019930923357605934, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00012936958228237927, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00013, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0023675146512687206, + "learning_rate": 4.337651127546448e-05, + "loss": 2.4802844563964754e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0036730507854372263, + "learning_rate": 4.3096315242201736e-05, + "loss": 3.144190486636944e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0003704871633090079, + "learning_rate": 4.2816614569850635e-05, + "loss": 8.931780030252412e-06, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.43, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.05240330100059509, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00026115920627489686, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00026, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.0012482519960030913, + "learning_rate": 4.225874584473174e-05, + "loss": 2.5392351744812913e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.050520945340394974, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0006884933682158589, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00069, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.002656160155311227, + "learning_rate": 4.170299795992081e-05, + "loss": 2.599508843559306e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.19733326137065887, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0016555915353819728, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00166, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.008795367553830147, + "learning_rate": 4.114946342220728e-05, + "loss": 6.713026959914714e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.00049841299187392, + "learning_rate": 4.087355496656321e-05, + "loss": 1.1344342965458054e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.23748824000358582, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0030270391143858433, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00303, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.19, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0024427378084510565, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.495265627861954e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00003, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.01809818297624588, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00013907425454817712, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0010443986393511295, + "learning_rate": 3.977591418134619e-05, + "loss": 1.951586818904616e-05, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.009523316286504269, + "learning_rate": 3.95030593412612e-05, + "loss": 6.634376040892676e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00007, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.049099843949079514, + "learning_rate": 3.923084939213296e-05, + "loss": 0.00039430244942195714, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 32.11, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.01699669286608696, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00013168745499569923, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00013, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.07120174169540405, + "learning_rate": 3.868840945050728e-05, + "loss": 0.000350497430190444, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00035, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.001589596620760858, + "learning_rate": 3.841820203114995e-05, + "loss": 2.160290750907734e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.014652811922132969, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00011893574264831841, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.032810863107442856, + "learning_rate": 3.787986851704667e-05, + "loss": 0.00019718095427379012, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.02765781246125698, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00016869997489266098, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00017, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.01766805537045002, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00014554498193319887, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00015, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.05623512342572212, + "learning_rate": 3.707773935267552e-05, + "loss": 0.00024310205481015146, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00024, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.0034802353475242853, + "learning_rate": 3.68118397962661e-05, + "loss": 3.652479063021019e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 31.03, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.0005486137815751135, + "learning_rate": 3.654669712344384e-05, + "loss": 1.2760298886860255e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.053988587111234665, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004493523156270385, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00045, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.61, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.023256463930010796, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0001921930379467085, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00019, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.87, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.005770122166723013, + "learning_rate": 3.575592058295017e-05, + "loss": 5.241332473815419e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0003110536199528724, + "learning_rate": 3.549391545931585e-05, + "loss": 5.897365099372109e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.008365475572645664, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.736510451650247e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00008, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.003012418746948242, + "learning_rate": 3.49723512647657e-05, + "loss": 3.9521881262771785e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.00036992691457271576, + "learning_rate": 3.471281389826491e-05, + "loss": 9.513973964203615e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.77, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.004851094912737608, + "learning_rate": 3.4454120764386764e-05, + "loss": 3.941698741982691e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.08602030575275421, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.0006739848759025335, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00067, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.007851852104067802, + "learning_rate": 3.3939310220027456e-05, + "loss": 7.82167335273698e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00008, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.34, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.000522423128131777, + "learning_rate": 3.3683214232914404e-05, + "loss": 7.171449851739453e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00001, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0008156000403687358, + "learning_rate": 3.342800532426873e-05, + "loss": 1.2959781088284217e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.012755601666867733, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00010119444777956232, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0001, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.0019102037185803056, + "learning_rate": 3.292029118616024e-05, + "loss": 2.8528616894618608e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.07546474039554596, + "learning_rate": 3.266780708475511e-05, + "loss": 0.0005027748411521316, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0005, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.014789975248277187, + "learning_rate": 3.241625231705354e-05, + "loss": 0.00011153960804222152, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.001108153141103685, + "learning_rate": 3.216563735127618e-05, + "loss": 1.160802275990136e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.061208564788103104, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0005815322510898113, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00058, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.0031761936843395233, + "learning_rate": 3.166726850239794e-05, + "loss": 2.453591878293082e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0003757224476430565, + "learning_rate": 3.141953535845912e-05, + "loss": 6.875088274682639e-06, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0009734284249134362, + "learning_rate": 3.11727834939056e-05, + "loss": 9.81441462499788e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.0010252115316689014, + "learning_rate": 3.092702317708967e-05, + "loss": 8.114137017400935e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.0002301523054484278, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.338565642887261e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00001, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.0006164724472910166, + "learning_rate": 3.0438518053342407e-05, + "loss": 8.090713890851475e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0006751597975380719, + "learning_rate": 3.0195793575103266e-05, + "loss": 1.1418029316700995e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.00013061391655355692, + "learning_rate": 2.9954101301140146e-05, + "loss": 3.586079174056067e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.001630443730391562, + "learning_rate": 2.9713451289255123e-05, + "loss": 1.7084981664083898e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00002, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0003005889884661883, + "learning_rate": 2.9473853553877484e-05, + "loss": 6.793928150727879e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.005686561111360788, + "learning_rate": 2.9235318065647e-05, + "loss": 5.100644557387568e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.43, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.0014047786826267838, + "learning_rate": 2.8997854750998964e-05, + "loss": 9.773429155757185e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 9.058098657988012e-05, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.382128286626539e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.0, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.0009709852165542543, + "learning_rate": 2.8526184124692883e-05, + "loss": 7.577294582006289e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.89, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.0034374946262687445, + "learning_rate": 2.829199644117484e-05, + "loss": 3.723455301951617e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.0033002276904881, + "learning_rate": 2.8058920186702553e-05, + "loss": 2.274679354741238e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.0006988391978666186, + "learning_rate": 2.782696506053033e-05, + "loss": 5.0486655709391925e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.0052696517668664455, + "learning_rate": 2.7596140715257824e-05, + "loss": 2.7391106414142996e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00003, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.0015628528781235218, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.4505098079098389e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.0021482163574546576, + "learning_rate": 2.7137922742128486e-05, + "loss": 1.399279062752612e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.001900387927889824, + "learning_rate": 2.691054818259188e-05, + "loss": 1.4173308954923414e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 35.04, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.003817511023953557, + "learning_rate": 2.6684342539801933e-05, + "loss": 2.5515182642266154e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00003, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.3609258234500885, + "learning_rate": 2.645931522709877e-05, + "loss": 0.006828296929597855, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00685, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 41.02, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.0038110739551484585, + "learning_rate": 2.6235475608787365e-05, + "loss": 3.2214695238508284e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0038299988955259323, + "learning_rate": 2.6012832999747916e-05, + "loss": 1.8610287952469662e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00002, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.24727243185043335, + "learning_rate": 2.579139666504821e-05, + "loss": 0.029269637539982796, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0297, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.08204144984483719, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0006305916467681527, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00063, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.016210485249757767, + "learning_rate": 2.5352179627565532e-05, + "loss": 9.964105993276462e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0001, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.99, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.00014741058112122118, + "learning_rate": 2.5134417202396277e-05, + "loss": 4.461833668756299e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.00043298961827531457, + "learning_rate": 2.491789760603361e-05, + "loss": 1.0579917216091417e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.0009353766799904406, + "learning_rate": 2.4702629848741764e-05, + "loss": 1.3036698874202557e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.004666164051741362, + "learning_rate": 2.4488622888690785e-05, + "loss": 5.6815359130268916e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.0007844759384170175, + "learning_rate": 2.427588563158384e-05, + "loss": 1.926398545037955e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0016678960528224707, + "learning_rate": 2.406442693028651e-05, + "loss": 2.7389756724005565e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.00107457279227674, + "learning_rate": 2.3854255584458547e-05, + "loss": 2.128102642018348e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.0008889543823897839, + "learning_rate": 2.3645380340187508e-05, + "loss": 1.6756177501520142e-05, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.09012716263532639, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0013020217884331942, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0013, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.02159537374973297, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00023354997392743826, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00023, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.015429600141942501, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.00018487832858227193, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.004461898934096098, + "learning_rate": 2.2823013405081507e-05, + "loss": 4.71106031909585e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 39.0, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.0018961423775181174, + "learning_rate": 2.2620747959533722e-05, + "loss": 4.23475939896889e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00004, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.0012804885627701879, + "learning_rate": 2.2419829946830123e-05, + "loss": 2.206004865001887e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.0014976069796830416, + "learning_rate": 2.2220267727989325e-05, + "loss": 2.556692015787121e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.005655917339026928, + "learning_rate": 2.202206960760984e-05, + "loss": 5.6497796322219074e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00006, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.01994832046329975, + "learning_rate": 2.182524383352446e-05, + "loss": 0.00019567189156077802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0002, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.004994136746972799, + "learning_rate": 2.1629798596457056e-05, + "loss": 4.0259299566969275e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.0004713181406259537, + "learning_rate": 2.1435742029681725e-05, + "loss": 1.3051702808297705e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00034305796725675464, + "learning_rate": 2.124308220868431e-05, + "loss": 1.262360456166789e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.15386679768562317, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0018648281693458557, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00187, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.003915456123650074, + "learning_rate": 2.0861984815011552e-05, + "loss": 4.141925091971643e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.0021876280661672354, + "learning_rate": 2.0673563101354323e-05, + "loss": 2.675295399967581e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.0006817729445174336, + "learning_rate": 2.0486569850851317e-05, + "loss": 1.70326584338909e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0010018590837717056, + "learning_rate": 2.0301012845054956e-05, + "loss": 2.3183692974271253e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.89, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.001717094099149108, + "learning_rate": 2.011689980574966e-05, + "loss": 1.955918014573399e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.0029396964237093925, + "learning_rate": 1.993423839463052e-05, + "loss": 3.9509199268650264e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.3, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.0014759147306904197, + "learning_rate": 1.975303621298445e-05, + "loss": 1.9150986190652475e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.002330441027879715, + "learning_rate": 1.957330080137385e-05, + "loss": 3.6902769352309406e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.0019221253460273147, + "learning_rate": 1.9395039639322864e-05, + "loss": 3.196115358150564e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.17172899842262268, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0011913730995729566, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00119, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.006395282223820686, + "learning_rate": 1.904296967493982e-05, + "loss": 8.003453694982454e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.03568729758262634, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003129084361717105, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00031, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0005958312540315092, + "learning_rate": 1.869688492349885e-05, + "loss": 1.7803757145884447e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.04430711269378662, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0004007165553048253, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0004, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 190306 + } + ], + "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": 8.546068293579556e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-416/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-448/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-448/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0e1d99c17fb02d3d2d013e308d47fdeff89560fe --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ddab61d8202f7c1b2b90f50b916d2190d69acc0de38112a102ce9928746082e7 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-448/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..338c39db6ac59124868f07f5445daa5bc7b27363 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fad8c2740aa902c6e9c7743ee80e28ea997db6d411814062b88f420ea7e5a07f +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..1329b266df9f032a7779db44cbe92f7110ae5a5a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d27d98c90c4450806b1ea70f841a527e6e8f8e0dd5e5e1af25cbd5dd7470342a +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2045249a1135f9f2d87c87c3e02e6227697d0cfa --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fa78a58752cdd4bcff72584e7054dccc1d7ab6d6 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/tokens_state.json @@ -0,0 +1 @@ +{"total": 13560080, "trainable": 204821} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..85034bc7e416943758184b68c0c5e3561dbf2d2a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.004979890305548906, + "learning_rate": 4.9354244499126866e-05, + "loss": 3.968320743297227e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0025452408008277416, + "learning_rate": 4.90663667927174e-05, + "loss": 3.7232457543723285e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00004, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0009856978431344032, + "learning_rate": 4.877873600900581e-05, + "loss": 2.570721335359849e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0006475941045209765, + "learning_rate": 4.849136411748306e-05, + "loss": 1.7373771697748452e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.2577264606952667, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0018721828237175941, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00187, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.93, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.09887456148862839, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0007890360429883003, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00079, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.005742037668824196, + "learning_rate": 4.7630921326368736e-05, + "loss": 5.973849329166114e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.0062770843505859375, + "learning_rate": 4.7344704475577916e-05, + "loss": 7.13710323907435e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.009947159327566624, + "learning_rate": 4.705880619288153e-05, + "loss": 0.00010848060628632084, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.41, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.006360637955367565, + "learning_rate": 4.677323837567412e-05, + "loss": 5.382625022321008e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.030510911718010902, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00030309066642075777, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0003, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.004953032359480858, + "learning_rate": 4.620314165804964e-05, + "loss": 5.149709977558814e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.003854556707665324, + "learning_rate": 4.591863648168407e-05, + "loss": 4.885891394224018e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.11261430382728577, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0007564757252112031, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00076, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.03299031779170036, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00041300299926660955, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00041, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.0065305293537676334, + "learning_rate": 4.506743570677743e-05, + "loss": 7.41428229957819e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.005171677563339472, + "learning_rate": 4.478451305763618e-05, + "loss": 6.253476749407127e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.006580202840268612, + "learning_rate": 4.450201551660454e-05, + "loss": 5.272469570627436e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00005, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.31961339712142944, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.003208739683032036, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00321, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.0023711167741566896, + "learning_rate": 4.393834276419352e-05, + "loss": 3.434348400332965e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.019930923357605934, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00012936958228237927, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00013, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0023675146512687206, + "learning_rate": 4.337651127546448e-05, + "loss": 2.4802844563964754e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0036730507854372263, + "learning_rate": 4.3096315242201736e-05, + "loss": 3.144190486636944e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0003704871633090079, + "learning_rate": 4.2816614569850635e-05, + "loss": 8.931780030252412e-06, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.43, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.05240330100059509, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00026115920627489686, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00026, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.0012482519960030913, + "learning_rate": 4.225874584473174e-05, + "loss": 2.5392351744812913e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.050520945340394974, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0006884933682158589, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00069, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.002656160155311227, + "learning_rate": 4.170299795992081e-05, + "loss": 2.599508843559306e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.19733326137065887, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0016555915353819728, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00166, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.008795367553830147, + "learning_rate": 4.114946342220728e-05, + "loss": 6.713026959914714e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.00049841299187392, + "learning_rate": 4.087355496656321e-05, + "loss": 1.1344342965458054e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.23748824000358582, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0030270391143858433, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00303, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.19, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0024427378084510565, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.495265627861954e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00003, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.01809818297624588, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00013907425454817712, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0010443986393511295, + "learning_rate": 3.977591418134619e-05, + "loss": 1.951586818904616e-05, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.009523316286504269, + "learning_rate": 3.95030593412612e-05, + "loss": 6.634376040892676e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00007, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.049099843949079514, + "learning_rate": 3.923084939213296e-05, + "loss": 0.00039430244942195714, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 32.11, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.01699669286608696, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00013168745499569923, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00013, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.07120174169540405, + "learning_rate": 3.868840945050728e-05, + "loss": 0.000350497430190444, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00035, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.001589596620760858, + "learning_rate": 3.841820203114995e-05, + "loss": 2.160290750907734e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.014652811922132969, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00011893574264831841, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.032810863107442856, + "learning_rate": 3.787986851704667e-05, + "loss": 0.00019718095427379012, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.02765781246125698, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00016869997489266098, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00017, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.01766805537045002, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00014554498193319887, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00015, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.05623512342572212, + "learning_rate": 3.707773935267552e-05, + "loss": 0.00024310205481015146, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00024, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.0034802353475242853, + "learning_rate": 3.68118397962661e-05, + "loss": 3.652479063021019e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 31.03, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.0005486137815751135, + "learning_rate": 3.654669712344384e-05, + "loss": 1.2760298886860255e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.053988587111234665, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004493523156270385, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00045, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.61, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.023256463930010796, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0001921930379467085, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00019, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.87, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.005770122166723013, + "learning_rate": 3.575592058295017e-05, + "loss": 5.241332473815419e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0003110536199528724, + "learning_rate": 3.549391545931585e-05, + "loss": 5.897365099372109e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.008365475572645664, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.736510451650247e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00008, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.003012418746948242, + "learning_rate": 3.49723512647657e-05, + "loss": 3.9521881262771785e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.00036992691457271576, + "learning_rate": 3.471281389826491e-05, + "loss": 9.513973964203615e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.77, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.004851094912737608, + "learning_rate": 3.4454120764386764e-05, + "loss": 3.941698741982691e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.08602030575275421, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.0006739848759025335, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00067, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.007851852104067802, + "learning_rate": 3.3939310220027456e-05, + "loss": 7.82167335273698e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00008, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.34, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.000522423128131777, + "learning_rate": 3.3683214232914404e-05, + "loss": 7.171449851739453e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00001, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0008156000403687358, + "learning_rate": 3.342800532426873e-05, + "loss": 1.2959781088284217e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.012755601666867733, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00010119444777956232, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0001, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.0019102037185803056, + "learning_rate": 3.292029118616024e-05, + "loss": 2.8528616894618608e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.07546474039554596, + "learning_rate": 3.266780708475511e-05, + "loss": 0.0005027748411521316, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0005, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.014789975248277187, + "learning_rate": 3.241625231705354e-05, + "loss": 0.00011153960804222152, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.001108153141103685, + "learning_rate": 3.216563735127618e-05, + "loss": 1.160802275990136e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.061208564788103104, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0005815322510898113, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00058, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.0031761936843395233, + "learning_rate": 3.166726850239794e-05, + "loss": 2.453591878293082e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0003757224476430565, + "learning_rate": 3.141953535845912e-05, + "loss": 6.875088274682639e-06, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0009734284249134362, + "learning_rate": 3.11727834939056e-05, + "loss": 9.81441462499788e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.0010252115316689014, + "learning_rate": 3.092702317708967e-05, + "loss": 8.114137017400935e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.0002301523054484278, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.338565642887261e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00001, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.0006164724472910166, + "learning_rate": 3.0438518053342407e-05, + "loss": 8.090713890851475e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0006751597975380719, + "learning_rate": 3.0195793575103266e-05, + "loss": 1.1418029316700995e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.00013061391655355692, + "learning_rate": 2.9954101301140146e-05, + "loss": 3.586079174056067e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.001630443730391562, + "learning_rate": 2.9713451289255123e-05, + "loss": 1.7084981664083898e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00002, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0003005889884661883, + "learning_rate": 2.9473853553877484e-05, + "loss": 6.793928150727879e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.005686561111360788, + "learning_rate": 2.9235318065647e-05, + "loss": 5.100644557387568e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.43, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.0014047786826267838, + "learning_rate": 2.8997854750998964e-05, + "loss": 9.773429155757185e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 9.058098657988012e-05, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.382128286626539e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.0, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.0009709852165542543, + "learning_rate": 2.8526184124692883e-05, + "loss": 7.577294582006289e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.89, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.0034374946262687445, + "learning_rate": 2.829199644117484e-05, + "loss": 3.723455301951617e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.0033002276904881, + "learning_rate": 2.8058920186702553e-05, + "loss": 2.274679354741238e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.0006988391978666186, + "learning_rate": 2.782696506053033e-05, + "loss": 5.0486655709391925e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.0052696517668664455, + "learning_rate": 2.7596140715257824e-05, + "loss": 2.7391106414142996e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00003, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.0015628528781235218, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.4505098079098389e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.0021482163574546576, + "learning_rate": 2.7137922742128486e-05, + "loss": 1.399279062752612e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.001900387927889824, + "learning_rate": 2.691054818259188e-05, + "loss": 1.4173308954923414e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 35.04, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.003817511023953557, + "learning_rate": 2.6684342539801933e-05, + "loss": 2.5515182642266154e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00003, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.3609258234500885, + "learning_rate": 2.645931522709877e-05, + "loss": 0.006828296929597855, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00685, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 41.02, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.0038110739551484585, + "learning_rate": 2.6235475608787365e-05, + "loss": 3.2214695238508284e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0038299988955259323, + "learning_rate": 2.6012832999747916e-05, + "loss": 1.8610287952469662e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00002, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.24727243185043335, + "learning_rate": 2.579139666504821e-05, + "loss": 0.029269637539982796, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0297, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.08204144984483719, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0006305916467681527, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00063, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.016210485249757767, + "learning_rate": 2.5352179627565532e-05, + "loss": 9.964105993276462e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0001, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.99, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.00014741058112122118, + "learning_rate": 2.5134417202396277e-05, + "loss": 4.461833668756299e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.00043298961827531457, + "learning_rate": 2.491789760603361e-05, + "loss": 1.0579917216091417e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.0009353766799904406, + "learning_rate": 2.4702629848741764e-05, + "loss": 1.3036698874202557e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.004666164051741362, + "learning_rate": 2.4488622888690785e-05, + "loss": 5.6815359130268916e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.0007844759384170175, + "learning_rate": 2.427588563158384e-05, + "loss": 1.926398545037955e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0016678960528224707, + "learning_rate": 2.406442693028651e-05, + "loss": 2.7389756724005565e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.00107457279227674, + "learning_rate": 2.3854255584458547e-05, + "loss": 2.128102642018348e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.0008889543823897839, + "learning_rate": 2.3645380340187508e-05, + "loss": 1.6756177501520142e-05, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.09012716263532639, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0013020217884331942, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0013, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.02159537374973297, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00023354997392743826, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00023, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.015429600141942501, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.00018487832858227193, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.004461898934096098, + "learning_rate": 2.2823013405081507e-05, + "loss": 4.71106031909585e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 39.0, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.0018961423775181174, + "learning_rate": 2.2620747959533722e-05, + "loss": 4.23475939896889e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00004, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.0012804885627701879, + "learning_rate": 2.2419829946830123e-05, + "loss": 2.206004865001887e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.0014976069796830416, + "learning_rate": 2.2220267727989325e-05, + "loss": 2.556692015787121e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.005655917339026928, + "learning_rate": 2.202206960760984e-05, + "loss": 5.6497796322219074e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00006, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.01994832046329975, + "learning_rate": 2.182524383352446e-05, + "loss": 0.00019567189156077802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0002, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.004994136746972799, + "learning_rate": 2.1629798596457056e-05, + "loss": 4.0259299566969275e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.0004713181406259537, + "learning_rate": 2.1435742029681725e-05, + "loss": 1.3051702808297705e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00034305796725675464, + "learning_rate": 2.124308220868431e-05, + "loss": 1.262360456166789e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.15386679768562317, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0018648281693458557, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00187, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.003915456123650074, + "learning_rate": 2.0861984815011552e-05, + "loss": 4.141925091971643e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.0021876280661672354, + "learning_rate": 2.0673563101354323e-05, + "loss": 2.675295399967581e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.0006817729445174336, + "learning_rate": 2.0486569850851317e-05, + "loss": 1.70326584338909e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0010018590837717056, + "learning_rate": 2.0301012845054956e-05, + "loss": 2.3183692974271253e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.89, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.001717094099149108, + "learning_rate": 2.011689980574966e-05, + "loss": 1.955918014573399e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.0029396964237093925, + "learning_rate": 1.993423839463052e-05, + "loss": 3.9509199268650264e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.3, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.0014759147306904197, + "learning_rate": 1.975303621298445e-05, + "loss": 1.9150986190652475e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.002330441027879715, + "learning_rate": 1.957330080137385e-05, + "loss": 3.6902769352309406e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.0019221253460273147, + "learning_rate": 1.9395039639322864e-05, + "loss": 3.196115358150564e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.17172899842262268, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0011913730995729566, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00119, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.006395282223820686, + "learning_rate": 1.904296967493982e-05, + "loss": 8.003453694982454e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.03568729758262634, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003129084361717105, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00031, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0005958312540315092, + "learning_rate": 1.869688492349885e-05, + "loss": 1.7803757145884447e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.04430711269378662, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0004007165553048253, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0004, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0005576324183493853, + "learning_rate": 1.8356842992397304e-05, + "loss": 1.260453700524522e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.0026761293411254883, + "learning_rate": 1.8189105812005714e-05, + "loss": 5.3298615966923535e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.03599291294813156, + "learning_rate": 1.802290048317732e-05, + "loss": 0.0003095666761510074, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00031, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.0009753056219778955, + "learning_rate": 1.785823392239424e-05, + "loss": 2.4977898647193797e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.01631380431354046, + "learning_rate": 1.7695112982104225e-05, + "loss": 8.942006388679147e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.0004339451261330396, + "learning_rate": 1.7533544450435433e-05, + "loss": 1.288170551561052e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00001, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.003464407753199339, + "learning_rate": 1.7373535050913946e-05, + "loss": 2.005351780098863e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00002, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.25, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.020527612417936325, + "learning_rate": 1.721509144218405e-05, + "loss": 0.00022681039990857244, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.005457657855004072, + "learning_rate": 1.705822021773101e-05, + "loss": 5.116407555760816e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.12875990569591522, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0006953739793971181, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0007, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.00037059199530631304, + "learning_rate": 1.6749220968158415e-05, + "loss": 1.1453635124780703e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 39.27, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0024775040801614523, + "learning_rate": 1.659710580175893e-05, + "loss": 3.973771890741773e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 37.07, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.047215159982442856, + "learning_rate": 1.644658873654133e-05, + "loss": 0.0003503131156321615, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 429, + "tokens/total": 12985328, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.00022129464196041226, + "learning_rate": 1.629767603613508e-05, + "loss": 7.348043709498597e-06, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00001, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.036828842014074326, + "learning_rate": 1.615037389740547e-05, + "loss": 0.00021110885427333415, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00021, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.35, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.04464258998632431, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0003122977796010673, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00031, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.00044109905138611794, + "learning_rate": 1.5860625757072092e-05, + "loss": 9.314219823863823e-06, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 433, + "tokens/total": 13106128, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.0008370514842681587, + "learning_rate": 1.571819181307116e-05, + "loss": 1.3771925296168774e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.0005812773015350103, + "learning_rate": 1.557739254545075e-05, + "loss": 1.5841713320696726e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.002782998373731971, + "learning_rate": 1.543823381344311e-05, + "loss": 5.194370896788314e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00005, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.00162645208183676, + "learning_rate": 1.5300721408011114e-05, + "loss": 2.6176341634709388e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.004371184855699539, + "learning_rate": 1.5164861051607254e-05, + "loss": 4.292160156182945e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.004319014493376017, + "learning_rate": 1.5030658397935521e-05, + "loss": 4.223985888529569e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.0004699853016063571, + "learning_rate": 1.4898119031716104e-05, + "loss": 1.3975365618534852e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.00024973336257971823, + "learning_rate": 1.476724846845306e-05, + "loss": 8.745013474253938e-06, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 441, + "tokens/total": 13349392, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.005719303153455257, + "learning_rate": 1.463805215420471e-05, + "loss": 5.815384065499529e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00006, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.001514942734502256, + "learning_rate": 1.451053546535705e-05, + "loss": 1.5278332284651697e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 36.03, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.7005385160446167, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0004397963348310441, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00044, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.006485369987785816, + "learning_rate": 1.4260562119706606e-05, + "loss": 4.91196769871749e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.8, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.0012794778449460864, + "learning_rate": 1.413811586531508e-05, + "loss": 1.9917219105991535e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.0035426984541118145, + "learning_rate": 1.4017370040713884e-05, + "loss": 2.9128377718734555e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.23535850644111633, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.0006318427622318268, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00063, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 204821 + } + ], + "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": 9.204018712361395e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-448/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-480/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-480/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..085534ff1eb7bab072aa4bcd82458320bc77927c --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75141177e9bccdf92b1ac95302b87d05dd67560b0535f7f736eead303fe25a9f +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-480/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..f5e6d59b6a7246e0f075d0fcc3873e767fd1eb98 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2b7c832be1bd63d2b26b4e654c51bf44f5012f24a16ebbaacac4825dba48a72 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..9da896c9d8f84289d77b53501f8d2b2d3e8bc3ec --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:106674c12c31685485a283c8c514a41fe51a83a990a0c55b807abd9181094cbe +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce1c4e63a6c6f8be8c137d3add4eda184d59a043 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3da71c55273035b5e1d531117df684eced67a866 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/tokens_state.json @@ -0,0 +1 @@ +{"total": 14531200, "trainable": 219412} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..68d40df508e135e60bf90c78a5cb88f7c573bac9 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.004979890305548906, + "learning_rate": 4.9354244499126866e-05, + "loss": 3.968320743297227e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0025452408008277416, + "learning_rate": 4.90663667927174e-05, + "loss": 3.7232457543723285e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00004, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0009856978431344032, + "learning_rate": 4.877873600900581e-05, + "loss": 2.570721335359849e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0006475941045209765, + "learning_rate": 4.849136411748306e-05, + "loss": 1.7373771697748452e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.2577264606952667, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0018721828237175941, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00187, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.93, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.09887456148862839, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0007890360429883003, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00079, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.005742037668824196, + "learning_rate": 4.7630921326368736e-05, + "loss": 5.973849329166114e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.0062770843505859375, + "learning_rate": 4.7344704475577916e-05, + "loss": 7.13710323907435e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.009947159327566624, + "learning_rate": 4.705880619288153e-05, + "loss": 0.00010848060628632084, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.41, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.006360637955367565, + "learning_rate": 4.677323837567412e-05, + "loss": 5.382625022321008e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.030510911718010902, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00030309066642075777, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0003, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.004953032359480858, + "learning_rate": 4.620314165804964e-05, + "loss": 5.149709977558814e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.003854556707665324, + "learning_rate": 4.591863648168407e-05, + "loss": 4.885891394224018e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.11261430382728577, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0007564757252112031, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00076, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.03299031779170036, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00041300299926660955, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00041, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.0065305293537676334, + "learning_rate": 4.506743570677743e-05, + "loss": 7.41428229957819e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.005171677563339472, + "learning_rate": 4.478451305763618e-05, + "loss": 6.253476749407127e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.006580202840268612, + "learning_rate": 4.450201551660454e-05, + "loss": 5.272469570627436e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00005, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.31961339712142944, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.003208739683032036, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00321, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.0023711167741566896, + "learning_rate": 4.393834276419352e-05, + "loss": 3.434348400332965e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.019930923357605934, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00012936958228237927, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00013, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0023675146512687206, + "learning_rate": 4.337651127546448e-05, + "loss": 2.4802844563964754e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0036730507854372263, + "learning_rate": 4.3096315242201736e-05, + "loss": 3.144190486636944e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0003704871633090079, + "learning_rate": 4.2816614569850635e-05, + "loss": 8.931780030252412e-06, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.43, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.05240330100059509, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00026115920627489686, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00026, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.0012482519960030913, + "learning_rate": 4.225874584473174e-05, + "loss": 2.5392351744812913e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.050520945340394974, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0006884933682158589, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00069, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.002656160155311227, + "learning_rate": 4.170299795992081e-05, + "loss": 2.599508843559306e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.19733326137065887, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0016555915353819728, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00166, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.008795367553830147, + "learning_rate": 4.114946342220728e-05, + "loss": 6.713026959914714e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.00049841299187392, + "learning_rate": 4.087355496656321e-05, + "loss": 1.1344342965458054e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.23748824000358582, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0030270391143858433, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00303, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.19, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0024427378084510565, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.495265627861954e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00003, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.01809818297624588, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00013907425454817712, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0010443986393511295, + "learning_rate": 3.977591418134619e-05, + "loss": 1.951586818904616e-05, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.009523316286504269, + "learning_rate": 3.95030593412612e-05, + "loss": 6.634376040892676e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00007, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.049099843949079514, + "learning_rate": 3.923084939213296e-05, + "loss": 0.00039430244942195714, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 32.11, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.01699669286608696, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00013168745499569923, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00013, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.07120174169540405, + "learning_rate": 3.868840945050728e-05, + "loss": 0.000350497430190444, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00035, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.001589596620760858, + "learning_rate": 3.841820203114995e-05, + "loss": 2.160290750907734e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.014652811922132969, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00011893574264831841, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.032810863107442856, + "learning_rate": 3.787986851704667e-05, + "loss": 0.00019718095427379012, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.02765781246125698, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00016869997489266098, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00017, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.01766805537045002, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00014554498193319887, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00015, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.05623512342572212, + "learning_rate": 3.707773935267552e-05, + "loss": 0.00024310205481015146, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00024, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.0034802353475242853, + "learning_rate": 3.68118397962661e-05, + "loss": 3.652479063021019e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 31.03, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.0005486137815751135, + "learning_rate": 3.654669712344384e-05, + "loss": 1.2760298886860255e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.053988587111234665, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004493523156270385, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00045, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.61, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.023256463930010796, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0001921930379467085, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00019, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.87, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.005770122166723013, + "learning_rate": 3.575592058295017e-05, + "loss": 5.241332473815419e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0003110536199528724, + "learning_rate": 3.549391545931585e-05, + "loss": 5.897365099372109e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.008365475572645664, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.736510451650247e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00008, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.003012418746948242, + "learning_rate": 3.49723512647657e-05, + "loss": 3.9521881262771785e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.00036992691457271576, + "learning_rate": 3.471281389826491e-05, + "loss": 9.513973964203615e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.77, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.004851094912737608, + "learning_rate": 3.4454120764386764e-05, + "loss": 3.941698741982691e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.08602030575275421, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.0006739848759025335, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00067, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.007851852104067802, + "learning_rate": 3.3939310220027456e-05, + "loss": 7.82167335273698e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00008, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.34, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.000522423128131777, + "learning_rate": 3.3683214232914404e-05, + "loss": 7.171449851739453e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00001, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0008156000403687358, + "learning_rate": 3.342800532426873e-05, + "loss": 1.2959781088284217e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.012755601666867733, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00010119444777956232, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0001, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.0019102037185803056, + "learning_rate": 3.292029118616024e-05, + "loss": 2.8528616894618608e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.07546474039554596, + "learning_rate": 3.266780708475511e-05, + "loss": 0.0005027748411521316, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0005, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.014789975248277187, + "learning_rate": 3.241625231705354e-05, + "loss": 0.00011153960804222152, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.001108153141103685, + "learning_rate": 3.216563735127618e-05, + "loss": 1.160802275990136e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.061208564788103104, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0005815322510898113, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00058, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.0031761936843395233, + "learning_rate": 3.166726850239794e-05, + "loss": 2.453591878293082e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0003757224476430565, + "learning_rate": 3.141953535845912e-05, + "loss": 6.875088274682639e-06, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0009734284249134362, + "learning_rate": 3.11727834939056e-05, + "loss": 9.81441462499788e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.0010252115316689014, + "learning_rate": 3.092702317708967e-05, + "loss": 8.114137017400935e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.0002301523054484278, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.338565642887261e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00001, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.0006164724472910166, + "learning_rate": 3.0438518053342407e-05, + "loss": 8.090713890851475e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0006751597975380719, + "learning_rate": 3.0195793575103266e-05, + "loss": 1.1418029316700995e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.00013061391655355692, + "learning_rate": 2.9954101301140146e-05, + "loss": 3.586079174056067e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.001630443730391562, + "learning_rate": 2.9713451289255123e-05, + "loss": 1.7084981664083898e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00002, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0003005889884661883, + "learning_rate": 2.9473853553877484e-05, + "loss": 6.793928150727879e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.005686561111360788, + "learning_rate": 2.9235318065647e-05, + "loss": 5.100644557387568e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.43, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.0014047786826267838, + "learning_rate": 2.8997854750998964e-05, + "loss": 9.773429155757185e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 9.058098657988012e-05, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.382128286626539e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.0, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.0009709852165542543, + "learning_rate": 2.8526184124692883e-05, + "loss": 7.577294582006289e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.89, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.0034374946262687445, + "learning_rate": 2.829199644117484e-05, + "loss": 3.723455301951617e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.0033002276904881, + "learning_rate": 2.8058920186702553e-05, + "loss": 2.274679354741238e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.0006988391978666186, + "learning_rate": 2.782696506053033e-05, + "loss": 5.0486655709391925e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.0052696517668664455, + "learning_rate": 2.7596140715257824e-05, + "loss": 2.7391106414142996e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00003, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.0015628528781235218, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.4505098079098389e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.0021482163574546576, + "learning_rate": 2.7137922742128486e-05, + "loss": 1.399279062752612e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.001900387927889824, + "learning_rate": 2.691054818259188e-05, + "loss": 1.4173308954923414e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 35.04, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.003817511023953557, + "learning_rate": 2.6684342539801933e-05, + "loss": 2.5515182642266154e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00003, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.3609258234500885, + "learning_rate": 2.645931522709877e-05, + "loss": 0.006828296929597855, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00685, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 41.02, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.0038110739551484585, + "learning_rate": 2.6235475608787365e-05, + "loss": 3.2214695238508284e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0038299988955259323, + "learning_rate": 2.6012832999747916e-05, + "loss": 1.8610287952469662e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00002, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.24727243185043335, + "learning_rate": 2.579139666504821e-05, + "loss": 0.029269637539982796, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0297, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.08204144984483719, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0006305916467681527, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00063, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.016210485249757767, + "learning_rate": 2.5352179627565532e-05, + "loss": 9.964105993276462e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0001, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.99, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.00014741058112122118, + "learning_rate": 2.5134417202396277e-05, + "loss": 4.461833668756299e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.00043298961827531457, + "learning_rate": 2.491789760603361e-05, + "loss": 1.0579917216091417e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.0009353766799904406, + "learning_rate": 2.4702629848741764e-05, + "loss": 1.3036698874202557e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.004666164051741362, + "learning_rate": 2.4488622888690785e-05, + "loss": 5.6815359130268916e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.0007844759384170175, + "learning_rate": 2.427588563158384e-05, + "loss": 1.926398545037955e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0016678960528224707, + "learning_rate": 2.406442693028651e-05, + "loss": 2.7389756724005565e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.00107457279227674, + "learning_rate": 2.3854255584458547e-05, + "loss": 2.128102642018348e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.0008889543823897839, + "learning_rate": 2.3645380340187508e-05, + "loss": 1.6756177501520142e-05, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.09012716263532639, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0013020217884331942, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0013, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.02159537374973297, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00023354997392743826, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00023, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.015429600141942501, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.00018487832858227193, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.004461898934096098, + "learning_rate": 2.2823013405081507e-05, + "loss": 4.71106031909585e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 39.0, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.0018961423775181174, + "learning_rate": 2.2620747959533722e-05, + "loss": 4.23475939896889e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00004, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.0012804885627701879, + "learning_rate": 2.2419829946830123e-05, + "loss": 2.206004865001887e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.0014976069796830416, + "learning_rate": 2.2220267727989325e-05, + "loss": 2.556692015787121e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.005655917339026928, + "learning_rate": 2.202206960760984e-05, + "loss": 5.6497796322219074e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00006, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.01994832046329975, + "learning_rate": 2.182524383352446e-05, + "loss": 0.00019567189156077802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0002, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.004994136746972799, + "learning_rate": 2.1629798596457056e-05, + "loss": 4.0259299566969275e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.0004713181406259537, + "learning_rate": 2.1435742029681725e-05, + "loss": 1.3051702808297705e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00034305796725675464, + "learning_rate": 2.124308220868431e-05, + "loss": 1.262360456166789e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.15386679768562317, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0018648281693458557, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00187, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.003915456123650074, + "learning_rate": 2.0861984815011552e-05, + "loss": 4.141925091971643e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.0021876280661672354, + "learning_rate": 2.0673563101354323e-05, + "loss": 2.675295399967581e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.0006817729445174336, + "learning_rate": 2.0486569850851317e-05, + "loss": 1.70326584338909e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0010018590837717056, + "learning_rate": 2.0301012845054956e-05, + "loss": 2.3183692974271253e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.89, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.001717094099149108, + "learning_rate": 2.011689980574966e-05, + "loss": 1.955918014573399e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.0029396964237093925, + "learning_rate": 1.993423839463052e-05, + "loss": 3.9509199268650264e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.3, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.0014759147306904197, + "learning_rate": 1.975303621298445e-05, + "loss": 1.9150986190652475e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.002330441027879715, + "learning_rate": 1.957330080137385e-05, + "loss": 3.6902769352309406e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.0019221253460273147, + "learning_rate": 1.9395039639322864e-05, + "loss": 3.196115358150564e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.17172899842262268, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0011913730995729566, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00119, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.006395282223820686, + "learning_rate": 1.904296967493982e-05, + "loss": 8.003453694982454e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.03568729758262634, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003129084361717105, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00031, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0005958312540315092, + "learning_rate": 1.869688492349885e-05, + "loss": 1.7803757145884447e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.04430711269378662, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0004007165553048253, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0004, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0005576324183493853, + "learning_rate": 1.8356842992397304e-05, + "loss": 1.260453700524522e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.0026761293411254883, + "learning_rate": 1.8189105812005714e-05, + "loss": 5.3298615966923535e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.03599291294813156, + "learning_rate": 1.802290048317732e-05, + "loss": 0.0003095666761510074, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00031, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.0009753056219778955, + "learning_rate": 1.785823392239424e-05, + "loss": 2.4977898647193797e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.01631380431354046, + "learning_rate": 1.7695112982104225e-05, + "loss": 8.942006388679147e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.0004339451261330396, + "learning_rate": 1.7533544450435433e-05, + "loss": 1.288170551561052e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00001, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.003464407753199339, + "learning_rate": 1.7373535050913946e-05, + "loss": 2.005351780098863e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00002, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.25, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.020527612417936325, + "learning_rate": 1.721509144218405e-05, + "loss": 0.00022681039990857244, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.005457657855004072, + "learning_rate": 1.705822021773101e-05, + "loss": 5.116407555760816e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.12875990569591522, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0006953739793971181, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0007, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.00037059199530631304, + "learning_rate": 1.6749220968158415e-05, + "loss": 1.1453635124780703e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 39.27, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0024775040801614523, + "learning_rate": 1.659710580175893e-05, + "loss": 3.973771890741773e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 37.07, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.047215159982442856, + "learning_rate": 1.644658873654133e-05, + "loss": 0.0003503131156321615, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 429, + "tokens/total": 12985328, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.00022129464196041226, + "learning_rate": 1.629767603613508e-05, + "loss": 7.348043709498597e-06, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00001, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.036828842014074326, + "learning_rate": 1.615037389740547e-05, + "loss": 0.00021110885427333415, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00021, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.35, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.04464258998632431, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0003122977796010673, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00031, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.00044109905138611794, + "learning_rate": 1.5860625757072092e-05, + "loss": 9.314219823863823e-06, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 433, + "tokens/total": 13106128, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.0008370514842681587, + "learning_rate": 1.571819181307116e-05, + "loss": 1.3771925296168774e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.0005812773015350103, + "learning_rate": 1.557739254545075e-05, + "loss": 1.5841713320696726e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.002782998373731971, + "learning_rate": 1.543823381344311e-05, + "loss": 5.194370896788314e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00005, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.00162645208183676, + "learning_rate": 1.5300721408011114e-05, + "loss": 2.6176341634709388e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.004371184855699539, + "learning_rate": 1.5164861051607254e-05, + "loss": 4.292160156182945e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.004319014493376017, + "learning_rate": 1.5030658397935521e-05, + "loss": 4.223985888529569e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.0004699853016063571, + "learning_rate": 1.4898119031716104e-05, + "loss": 1.3975365618534852e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.00024973336257971823, + "learning_rate": 1.476724846845306e-05, + "loss": 8.745013474253938e-06, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 441, + "tokens/total": 13349392, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.005719303153455257, + "learning_rate": 1.463805215420471e-05, + "loss": 5.815384065499529e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00006, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.001514942734502256, + "learning_rate": 1.451053546535705e-05, + "loss": 1.5278332284651697e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 36.03, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.7005385160446167, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0004397963348310441, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00044, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.006485369987785816, + "learning_rate": 1.4260562119706606e-05, + "loss": 4.91196769871749e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.8, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.0012794778449460864, + "learning_rate": 1.413811586531508e-05, + "loss": 1.9917219105991535e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.0035426984541118145, + "learning_rate": 1.4017370040713884e-05, + "loss": 2.9128377718734555e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.23535850644111633, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.0006318427622318268, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00063, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 204821 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.0008378064376302063, + "learning_rate": 1.3780999708818058e-05, + "loss": 1.589788189448882e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 449, + "tokens/total": 13590496, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 205271 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.0006133823189884424, + "learning_rate": 1.3665385037857758e-05, + "loss": 1.3079279597150162e-05, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 450, + "tokens/total": 13620960, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 205723 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.002004537731409073, + "learning_rate": 1.3551490468947126e-05, + "loss": 2.534996201575268e-05, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 451, + "tokens/total": 13651344, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 206169 + }, + { + "epoch": 1.765625, + "grad_norm": 0.010971812531352043, + "learning_rate": 1.3439320741704075e-05, + "loss": 2.6202287699561566e-05, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 452, + "tokens/total": 13681792, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 206592 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.014560825191438198, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.00012574285210575908, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 453, + "tokens/total": 13712256, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 207055 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.007601496763527393, + "learning_rate": 1.3220174411609587e-05, + "loss": 4.6285953430924565e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00005, + "step": 454, + "tokens/total": 13742128, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 207494 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.0018729944713413715, + "learning_rate": 1.3113206928331471e-05, + "loss": 2.2211435862118378e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 455, + "tokens/total": 13772160, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 207956 + }, + { + "epoch": 1.78125, + "grad_norm": 0.010155079886317253, + "learning_rate": 1.300798252548806e-05, + "loss": 1.8203903891844675e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 456, + "tokens/total": 13802448, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 208408 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.004313099663704634, + "learning_rate": 1.2904505581896265e-05, + "loss": 3.647474659373984e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 457, + "tokens/total": 13832880, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 208889 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.000998936826363206, + "learning_rate": 1.2802780403654082e-05, + "loss": 2.4444412701996043e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00002, + "step": 458, + "tokens/total": 13862880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 209336 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.32933780550956726, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.00300136161968112, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00301, + "step": 459, + "tokens/total": 13893440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 209797 + }, + { + "epoch": 1.796875, + "grad_norm": 0.0005855838535353541, + "learning_rate": 1.2604602202943861e-05, + "loss": 1.4857862879580352e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 460, + "tokens/total": 13923824, + "tokens/train_per_sec_per_gpu": 36.62, + "tokens/trainable": 210240 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.020791945978999138, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.00013913628936279565, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00014, + "step": 461, + "tokens/total": 13954208, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 210702 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.010361927561461926, + "learning_rate": 1.2413480911029655e-05, + "loss": 5.078136018710211e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 462, + "tokens/total": 13984784, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 211140 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.0006120642647147179, + "learning_rate": 1.2320576593470082e-05, + "loss": 1.312442145717796e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00001, + "step": 463, + "tokens/total": 14014896, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 211575 + }, + { + "epoch": 1.8125, + "grad_norm": 0.008837698958814144, + "learning_rate": 1.2229448340928828e-05, + "loss": 8.223413169616833e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00008, + "step": 464, + "tokens/total": 14045312, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 212083 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.0006230191211216152, + "learning_rate": 1.2140099945624458e-05, + "loss": 7.008733518887311e-06, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 465, + "tokens/total": 14075840, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 212558 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.030772700905799866, + "learning_rate": 1.205253512570841e-05, + "loss": 0.00023961231636349112, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00024, + "step": 466, + "tokens/total": 14106016, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 213014 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0005919403047300875, + "learning_rate": 1.1966757525110255e-05, + "loss": 1.4494933566311374e-05, + "memory/device_reserved (GiB)": 37.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 467, + "tokens/total": 14136512, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 213465 + }, + { + "epoch": 1.828125, + "grad_norm": 0.012348134070634842, + "learning_rate": 1.1882770713386095e-05, + "loss": 9.488382784184068e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 468, + "tokens/total": 14166832, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 213940 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.03572385758161545, + "learning_rate": 1.180057818556998e-05, + "loss": 0.00021282717352733016, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 469, + "tokens/total": 14197360, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 214415 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.00024021990248002112, + "learning_rate": 1.1720183362028494e-05, + "loss": 7.366670615738258e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 470, + "tokens/total": 14227600, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 214880 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.050287000834941864, + "learning_rate": 1.1641589588318387e-05, + "loss": 0.00028066671802662313, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00028, + "step": 471, + "tokens/total": 14257888, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 215334 + }, + { + "epoch": 1.84375, + "grad_norm": 0.0002384333492955193, + "learning_rate": 1.1564800135047418e-05, + "loss": 7.073277629388031e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 472, + "tokens/total": 14288208, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 215822 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.00174198136664927, + "learning_rate": 1.148981819773816e-05, + "loss": 2.6067960789077915e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 473, + "tokens/total": 14318448, + "tokens/train_per_sec_per_gpu": 31.32, + "tokens/trainable": 216239 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.0004001902125310153, + "learning_rate": 1.1416646896695086e-05, + "loss": 8.549897756893188e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 474, + "tokens/total": 14348912, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 216691 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.001704781549051404, + "learning_rate": 1.1345289276874717e-05, + "loss": 1.559099655423779e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 475, + "tokens/total": 14379440, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 217162 + }, + { + "epoch": 1.859375, + "grad_norm": 0.004531325306743383, + "learning_rate": 1.1275748307758873e-05, + "loss": 4.107754648430273e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 476, + "tokens/total": 14409840, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 217638 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.009856121614575386, + "learning_rate": 1.1208026883231147e-05, + "loss": 2.674189818208106e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00003, + "step": 477, + "tokens/total": 14440256, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 218053 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.00023725698702037334, + "learning_rate": 1.1142127821456433e-05, + "loss": 7.282562819455052e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 478, + "tokens/total": 14470704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 218503 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.0008088731556199491, + "learning_rate": 1.1078053864763674e-05, + "loss": 1.4862998796161264e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 479, + "tokens/total": 14501072, + "tokens/train_per_sec_per_gpu": 36.66, + "tokens/trainable": 218964 + }, + { + "epoch": 1.875, + "grad_norm": 0.11135943233966827, + "learning_rate": 1.1015807679531756e-05, + "loss": 0.0015005860477685928, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0015, + "step": 480, + "tokens/total": 14531200, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 219412 + } + ], + "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": 9.863174606128128e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-480/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-512/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-512/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..32a98d0c896ae739414f6dad378cf34abb424d19 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82e5c558fd3518967c97a463cc23363818705003444ae831b5667d43e013cc03 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-512/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..abcdcb72d8518d6d68c5572bebd7bf7ae7b83bef --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:906d8543a4a822e6c81e78141cb51fe2165a634f54a2424c75faa38653b40c90 +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..0626d83bfee73afaca3ed469625b1cf5142c7a61 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b4b9d2fe88b5aebe9f6d0b2a3cb27b73b4b5baa17ae0760b0ccb0011a29ee6d +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..22325b3b3d08e0697784fe21b8c321485a51f7b5 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..115a5bf49aff4472c7ae208d22e2b298d69dbe1c --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokens_state.json @@ -0,0 +1 @@ +{"total": 15496368, "trainable": 234060} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..07873e2673f702c3d0255c956c73ad4b686d054b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.26704758405685425, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014278757385909557, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01438, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.15346744656562805, + "learning_rate": 9.523275153154695e-05, + "loss": 0.0042286161333322525, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00424, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.3834576904773712, + "learning_rate": 9.51018809682839e-05, + "loss": 0.011703035794198513, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01177, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.68, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.7086752653121948, + "learning_rate": 9.49693416020645e-05, + "loss": 0.014463641680777073, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01457, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.27034133672714233, + "learning_rate": 9.483513894839276e-05, + "loss": 0.00494391517713666, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00496, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.17598937451839447, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002658358309417963, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00266, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.22, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.25017818808555603, + "learning_rate": 9.456176618655689e-05, + "loss": 0.005640048533678055, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00566, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7971018552780151, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03405749425292015, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03464, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3616940677165985, + "learning_rate": 9.428180818692884e-05, + "loss": 0.004266355652362108, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00428, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.37930840253829956, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018286891281604767, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01846, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2938750088214874, + "learning_rate": 9.399531154980424e-05, + "loss": 0.007349148392677307, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00738, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.7292595505714417, + "learning_rate": 9.384962610259455e-05, + "loss": 0.008774684742093086, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00881, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.07, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.14338180422782898, + "learning_rate": 9.370232396386494e-05, + "loss": 0.003269340842962265, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00327, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.5673946142196655, + "learning_rate": 9.355341126345868e-05, + "loss": 0.015835193917155266, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01596, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4097978174686432, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015539250336587429, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01566, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2639133036136627, + "learning_rate": 9.325077903184159e-05, + "loss": 0.010360169224441051, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01041, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.03919512405991554, + "learning_rate": 9.30970720943932e-05, + "loss": 0.000984038575552404, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00098, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.517244815826416, + "learning_rate": 9.2941779782269e-05, + "loss": 0.014096877537667751, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0142, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.38755857944488525, + "learning_rate": 9.278490855781596e-05, + "loss": 0.010432498529553413, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01049, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.5933223366737366, + "learning_rate": 9.262646494908604e-05, + "loss": 0.018632538616657257, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01881, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 1.0803433656692505, + "learning_rate": 9.246645554956457e-05, + "loss": 0.018085651099681854, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01825, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.18590858578681946, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0024977584835141897, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0025, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 35.08, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.20355315506458282, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0031010895036160946, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00311, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.12187723815441132, + "learning_rate": 9.197709951682268e-05, + "loss": 0.0025250515900552273, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00253, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.4177272915840149, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01381619367748499, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01391, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.1338144689798355, + "learning_rate": 9.164315700760271e-05, + "loss": 0.002670921152457595, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00267, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.2588568329811096, + "learning_rate": 9.147389495587671e-05, + "loss": 0.005417007952928543, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00543, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.21248272061347961, + "learning_rate": 9.130311507650116e-05, + "loss": 0.002708941698074341, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00271, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.344331294298172, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00817079283297062, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0082, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.62, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.26439207792282104, + "learning_rate": 9.09570303250602e-05, + "loss": 0.005593656096607447, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00561, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.11266636103391647, + "learning_rate": 9.078173985499394e-05, + "loss": 0.0022880970500409603, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00229, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.11059004068374634, + "learning_rate": 9.060496036067713e-05, + "loss": 0.0021013133227825165, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0021, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.2581784129142761, + "learning_rate": 9.042669919862615e-05, + "loss": 0.010128125548362732, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01018, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2631431818008423, + "learning_rate": 9.024696378701557e-05, + "loss": 0.009796728380024433, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00984, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.08005515486001968, + "learning_rate": 9.006576160536948e-05, + "loss": 0.0013428543461486697, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00134, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.2250908762216568, + "learning_rate": 8.988310019425035e-05, + "loss": 0.0020651742815971375, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00207, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.25850772857666016, + "learning_rate": 8.969898715494506e-05, + "loss": 0.004448378924280405, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00446, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.4923039972782135, + "learning_rate": 8.951343014914869e-05, + "loss": 0.01576503925025463, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01589, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.019552618265151978, + "learning_rate": 8.932643689864568e-05, + "loss": 0.0003579487092792988, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.08227464556694031, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001130630262196064, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00113, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.4993005394935608, + "learning_rate": 8.894817284917364e-05, + "loss": 0.0084285419434309, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00846, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.10504501312971115, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0017363994847983122, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00174, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.34308046102523804, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0045009395107626915, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00451, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.4159857928752899, + "learning_rate": 8.837020140354295e-05, + "loss": 0.007854020223021507, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00788, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.6004462242126465, + "learning_rate": 8.817475616647554e-05, + "loss": 0.013760390691459179, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01386, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.011325858533382416, + "learning_rate": 8.797793039239017e-05, + "loss": 0.00024004181614145637, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00024, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.5004990100860596, + "learning_rate": 8.777973227201069e-05, + "loss": 0.01317769568413496, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01326, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.9903383255004883, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0297666247934103, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03021, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.21414855122566223, + "learning_rate": 8.737925204046629e-05, + "loss": 0.009566199965775013, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00961, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.43486288189888, + "learning_rate": 8.717698659491851e-05, + "loss": 0.006363986060023308, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00638, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4194156527519226, + "learning_rate": 8.697338213361735e-05, + "loss": 0.01065310649573803, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01071, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.1669190227985382, + "learning_rate": 8.676844712937552e-05, + "loss": 0.006577126681804657, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0066, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.88, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1995057761669159, + "learning_rate": 8.656219011037509e-05, + "loss": 0.006504375487565994, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00653, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.6449038982391357, + "learning_rate": 8.63546196598125e-05, + "loss": 0.020274966955184937, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02048, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.4044327437877655, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012972041964530945, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01306, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.45228511095046997, + "learning_rate": 8.593557306971349e-05, + "loss": 0.02865610271692276, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02907, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.1369956135749817, + "learning_rate": 8.572411436841618e-05, + "loss": 0.004469756968319416, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00448, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.132548987865448, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006225723773241043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00625, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.28, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.15946446359157562, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005463356152176857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00548, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.3434015214443207, + "learning_rate": 8.508210239396639e-05, + "loss": 0.014093048870563507, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01419, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.06986619532108307, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0025533996522426605, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00256, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.34366294741630554, + "learning_rate": 8.464782037243449e-05, + "loss": 0.013532277196645737, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01362, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.5110010504722595, + "learning_rate": 8.442882418044202e-05, + "loss": 0.01647217571735382, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01661, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.10558311641216278, + "learning_rate": 8.420860333495179e-05, + "loss": 0.004920906387269497, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00493, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.16564413905143738, + "learning_rate": 8.398716700025208e-05, + "loss": 0.00695970980450511, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00698, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.28341594338417053, + "learning_rate": 8.376452439121266e-05, + "loss": 0.010892837308347225, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01095, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.05444963648915291, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0015460889553651214, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00155, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.09271538257598877, + "learning_rate": 8.331565746019807e-05, + "loss": 0.0028422921895980835, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00285, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.20550787448883057, + "learning_rate": 8.308945181740812e-05, + "loss": 0.004469836596399546, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00448, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.3527662754058838, + "learning_rate": 8.286207725787153e-05, + "loss": 0.010943438857793808, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.011, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.4151509702205658, + "learning_rate": 8.263354324357182e-05, + "loss": 0.010490473359823227, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01055, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.13094273209571838, + "learning_rate": 8.240385928474219e-05, + "loss": 0.0015372417401522398, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00154, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.05917811021208763, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0017156600952148438, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00172, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.08870573341846466, + "learning_rate": 8.194107981329746e-05, + "loss": 0.0014357661129906774, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00144, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.16937553882598877, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0023305683862417936, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00233, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.68, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1098952665925026, + "learning_rate": 8.147381587530713e-05, + "loss": 0.001408248208463192, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00141, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20372426509857178, + "learning_rate": 8.123852650824877e-05, + "loss": 0.005307321436703205, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00532, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.97, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.19021031260490417, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0018209717236459255, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00182, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.03357760235667229, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0004772062529809773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.41071707010269165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.00923216249793768, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00927, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.1126730740070343, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0012899136636406183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00129, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.06933529675006866, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0005741716013289988, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00057, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.7, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.017655329778790474, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0002967978944070637, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0003, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.31013330817222595, + "learning_rate": 7.95614819466576e-05, + "loss": 0.0008732057176530361, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00087, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.4237620234489441, + "learning_rate": 7.931773536489872e-05, + "loss": 0.004198434762656689, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.00421, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.014504591003060341, + "learning_rate": 7.907297682291035e-05, + "loss": 0.00015536235878244042, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00016, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.0415632426738739, + "learning_rate": 7.882721650609442e-05, + "loss": 0.0004971292801201344, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0005, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.16170728206634521, + "learning_rate": 7.85804646415409e-05, + "loss": 0.0023218044079840183, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00232, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.014852911233901978, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0001844349899329245, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00018, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.4728831350803375, + "learning_rate": 7.808402738346527e-05, + "loss": 0.006746332161128521, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00677, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03623728081583977, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004019437765236944, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0004, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.2675865888595581, + "learning_rate": 7.758374768294647e-05, + "loss": 0.0020086662843823433, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.507493793964386, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0029892839957028627, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00299, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.0674358382821083, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0005840241792611778, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00058, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 1.451257348060608, + "learning_rate": 7.682630588562518e-05, + "loss": 0.010650807991623878, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01071, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.7005940079689026, + "learning_rate": 7.657199467573129e-05, + "loss": 0.012993288226425648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01308, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.08145342022180557, + "learning_rate": 7.631678576708561e-05, + "loss": 0.0008674697601236403, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.07144517451524734, + "learning_rate": 7.606068977997255e-05, + "loss": 0.000845041242428124, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.4820076823234558, + "learning_rate": 7.580371737159148e-05, + "loss": 0.00316348928026855, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00317, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.63, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.008089586161077023, + "learning_rate": 7.554587923561324e-05, + "loss": 0.00014078138337936252, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.4833170175552368, + "learning_rate": 7.528718610173511e-05, + "loss": 0.011923539452254772, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01199, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.4297773540019989, + "learning_rate": 7.502764873523431e-05, + "loss": 0.006487303879112005, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00651, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.30053746700286865, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0037448552902787924, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00375, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.5079417824745178, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01055203191936016, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01061, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3676057755947113, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00716105941683054, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00719, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.75, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.06194091588258743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.0004702930455096066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.10208877176046371, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0007544317631982267, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00075, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.014593830332159996, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00019093806622549891, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00019, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.21324966847896576, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0019866086076945066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.012929766438901424, + "learning_rate": 7.29222606473245e-05, + "loss": 0.00013661546108778566, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00014, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.48922187089920044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.014999452978372574, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01511, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.5681560039520264, + "learning_rate": 7.238823517544436e-05, + "loss": 0.011948533356189728, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01202, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 41.01, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.011107727885246277, + "learning_rate": 7.212013148295333e-05, + "loss": 0.00015826086746528745, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.044534292072057724, + "learning_rate": 7.185131535190975e-05, + "loss": 0.0009664896642789245, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00097, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.7783606648445129, + "learning_rate": 7.158179796885005e-05, + "loss": 0.02202792465686798, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.02227, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.17146360874176025, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0028479855973273516, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00285, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.19646775722503662, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00397948594763875, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00399, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.06095995381474495, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0011619608849287033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00116, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.13, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.08135152608156204, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0027059789281338453, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00271, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.14376476407051086, + "learning_rate": 7.022408581865382e-05, + "loss": 0.002945779822766781, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00295, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.030300427228212357, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0006251530721783638, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00063, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.08062481135129929, + "learning_rate": 6.967648691039213e-05, + "loss": 0.002811428625136614, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00282, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.21464644372463226, + "learning_rate": 6.940176563004123e-05, + "loss": 0.004250328987836838, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00426, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06321574747562408, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0014506221050396562, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00145, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.07537315040826797, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0018983217887580395, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0019, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.65, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.04364407807588577, + "learning_rate": 6.857405174478604e-05, + "loss": 0.0008961158455349505, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0009, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.28126901388168335, + "learning_rate": 6.82970020400792e-05, + "loss": 0.002431782428175211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00243, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.04008863866329193, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0008201852906495333, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00082, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.09057830274105072, + "learning_rate": 6.774125415526827e-05, + "loss": 0.00176344346255064, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00176, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.40913277864456177, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00800992175936699, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00804, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.06365697085857391, + "learning_rate": 6.718338543014937e-05, + "loss": 0.0008380536455661058, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00084, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.005372561980038881, + "learning_rate": 6.69036847577983e-05, + "loss": 8.878827793523669e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00009, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.62, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.3291122615337372, + "learning_rate": 6.662348872453553e-05, + "loss": 0.005115922540426254, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00513, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.99, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.005812515504658222, + "learning_rate": 6.63428089904618e-05, + "loss": 9.655929170548916e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.24782787263393402, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002415842143818736, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00242, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.04228518530726433, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0006638517370447516, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00066, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.24, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.13585986196994781, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0023324713110923767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00234, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.13806359469890594, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0027295195031911135, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00273, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.18551138043403625, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0018342758994549513, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00184, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.11544615775346756, + "learning_rate": 6.464922830953799e-05, + "loss": 0.002124304883182049, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00213, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.7617654800415039, + "learning_rate": 6.436549078207688e-05, + "loss": 0.014520522207021713, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01463, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.03847108036279678, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0005749939009547234, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00058, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.49, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.13679052889347076, + "learning_rate": 6.379685834195036e-05, + "loss": 0.003290478838607669, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0033, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.1627172827720642, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0017669250955805182, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00177, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.6043732166290283, + "learning_rate": 6.32267616243259e-05, + "loss": 0.015954118221998215, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01608, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16619658470153809, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0015951163368299603, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0016, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.94, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.024235516786575317, + "learning_rate": 6.265529552442209e-05, + "loss": 0.00045446690637618303, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00045, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.45, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.10136919468641281, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0024567395448684692, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00246, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.7829793691635132, + "learning_rate": 6.208255516539749e-05, + "loss": 0.004660450853407383, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00467, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.10806790739297867, + "learning_rate": 6.179573692313344e-05, + "loss": 0.001825700281187892, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00183, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.14372305572032928, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0018829500768333673, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00188, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.0468306727707386, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001063495990820229, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00106, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.56, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.3564869165420532, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.00227900268509984, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00228, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.02111588791012764, + "learning_rate": 6.064575550087316e-05, + "loss": 0.00027177060837857425, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00027, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.2769322395324707, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004786302801221609, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0048, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.12309696525335312, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.001197405974380672, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0012, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.23, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.6309447884559631, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.009977569803595543, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01003, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.022197438403964043, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0002314020530320704, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00023, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.13308732211589813, + "learning_rate": 5.920308275189541e-05, + "loss": 0.0008159374119713902, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00082, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.39, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.009844367392361164, + "learning_rate": 5.8913975409115874e-05, + "loss": 8.531998173566535e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.58, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.008356324397027493, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.252396219177172e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.005978915374726057, + "learning_rate": 5.833528413179249e-05, + "loss": 6.622956425417215e-05, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00007, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.82, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.12491501867771149, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0011766712414100766, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00118, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.45353081822395325, + "learning_rate": 5.77560376810977e-05, + "loss": 0.029483629390597343, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02992, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.02936917543411255, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0003550515975803137, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00036, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0044965120032429695, + "learning_rate": 5.717633247526522e-05, + "loss": 4.8715679440647364e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00005, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.62, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.11758256703615189, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0004940429353155196, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00049, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.22, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.003167251357808709, + "learning_rate": 5.659626500889066e-05, + "loss": 4.923924643662758e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00005, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0022706782910972834, + "learning_rate": 5.6306125599488905e-05, + "loss": 4.5954613597132266e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.2836518883705139, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0011350277345627546, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00114, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2555781304836273, + "learning_rate": 5.572569579717961e-05, + "loss": 0.0017866040579974651, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00179, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0076074316166341305, + "learning_rate": 5.543542955832538e-05, + "loss": 8.170487126335502e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.07716227322816849, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010607760632410645, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00106, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.45, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.018780827522277832, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0002637105353642255, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00026, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.03091493807733059, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0003143992798868567, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00031, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.13643087446689606, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016409320523962379, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00164, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 29.13, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.15425634384155273, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.0014877885114401579, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00149, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12966743111610413, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0022027853410691023, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00221, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.01629537157714367, + "learning_rate": 5.340373499110935e-05, + "loss": 0.00018065710901282728, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00018, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.14, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.010340728797018528, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00013175717322155833, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00013, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.12547697126865387, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0011799060739576817, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00118, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.013123616576194763, + "learning_rate": 5.2533763606737005e-05, + "loss": 8.206372876884416e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00008, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 30.09, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.02970624715089798, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00019968026026617736, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.0002, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.032595086842775345, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0004402882477734238, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00044, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.007255643606185913, + "learning_rate": 5.166471586820751e-05, + "loss": 7.451212150044739e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00007, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.25917166471481323, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.004500494804233313, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00451, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.016095004975795746, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.00017428149294573814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00017, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.26327428221702576, + "learning_rate": 5.079691724810461e-05, + "loss": 0.006136145442724228, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00616, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.08115784078836441, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0008479508687742054, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00085, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.013995068147778511, + "learning_rate": 5.021923930849237e-05, + "loss": 0.00012496854469645768, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00012, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.005962018389254808, + "learning_rate": 4.99306927511967e-05, + "loss": 5.7330969866598025e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00006, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.005454273894429207, + "learning_rate": 4.964235714846775e-05, + "loss": 6.549475801875815e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.63, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.004979890305548906, + "learning_rate": 4.9354244499126866e-05, + "loss": 3.968320743297227e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0025452408008277416, + "learning_rate": 4.90663667927174e-05, + "loss": 3.7232457543723285e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00004, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0009856978431344032, + "learning_rate": 4.877873600900581e-05, + "loss": 2.570721335359849e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.69, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0006475941045209765, + "learning_rate": 4.849136411748306e-05, + "loss": 1.7373771697748452e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.2577264606952667, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0018721828237175941, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00187, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.93, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.09887456148862839, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0007890360429883003, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00079, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 31.1, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.005742037668824196, + "learning_rate": 4.7630921326368736e-05, + "loss": 5.973849329166114e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.0062770843505859375, + "learning_rate": 4.7344704475577916e-05, + "loss": 7.13710323907435e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.009947159327566624, + "learning_rate": 4.705880619288153e-05, + "loss": 0.00010848060628632084, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.41, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.006360637955367565, + "learning_rate": 4.677323837567412e-05, + "loss": 5.382625022321008e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.030510911718010902, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00030309066642075777, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0003, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.004953032359480858, + "learning_rate": 4.620314165804964e-05, + "loss": 5.149709977558814e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.003854556707665324, + "learning_rate": 4.591863648168407e-05, + "loss": 4.885891394224018e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.84, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.11261430382728577, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0007564757252112031, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00076, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.03299031779170036, + "learning_rate": 4.535077169046201e-05, + "loss": 0.00041300299926660955, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00041, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.0065305293537676334, + "learning_rate": 4.506743570677743e-05, + "loss": 7.41428229957819e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.005171677563339472, + "learning_rate": 4.478451305763618e-05, + "loss": 6.253476749407127e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.57, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.006580202840268612, + "learning_rate": 4.450201551660454e-05, + "loss": 5.272469570627436e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00005, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.31961339712142944, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.003208739683032036, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00321, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.0023711167741566896, + "learning_rate": 4.393834276419352e-05, + "loss": 3.434348400332965e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.019930923357605934, + "learning_rate": 4.36571910095382e-05, + "loss": 0.00012936958228237927, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00013, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0023675146512687206, + "learning_rate": 4.337651127546448e-05, + "loss": 2.4802844563964754e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0036730507854372263, + "learning_rate": 4.3096315242201736e-05, + "loss": 3.144190486636944e-05, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0003704871633090079, + "learning_rate": 4.2816614569850635e-05, + "loss": 8.931780030252412e-06, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.43, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.05240330100059509, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.00026115920627489686, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00026, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.0012482519960030913, + "learning_rate": 4.225874584473174e-05, + "loss": 2.5392351744812913e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.050520945340394974, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0006884933682158589, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00069, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.002656160155311227, + "learning_rate": 4.170299795992081e-05, + "loss": 2.599508843559306e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.19733326137065887, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0016555915353819728, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00166, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 39.15, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.008795367553830147, + "learning_rate": 4.114946342220728e-05, + "loss": 6.713026959914714e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00007, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.00049841299187392, + "learning_rate": 4.087355496656321e-05, + "loss": 1.1344342965458054e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.23748824000358582, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0030270391143858433, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00303, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.19, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0024427378084510565, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.495265627861954e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00003, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.01809818297624588, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00013907425454817712, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0010443986393511295, + "learning_rate": 3.977591418134619e-05, + "loss": 1.951586818904616e-05, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.009523316286504269, + "learning_rate": 3.95030593412612e-05, + "loss": 6.634376040892676e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00007, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.049099843949079514, + "learning_rate": 3.923084939213296e-05, + "loss": 0.00039430244942195714, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 32.11, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.01699669286608696, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00013168745499569923, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00013, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.07120174169540405, + "learning_rate": 3.868840945050728e-05, + "loss": 0.000350497430190444, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00035, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.001589596620760858, + "learning_rate": 3.841820203114995e-05, + "loss": 2.160290750907734e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.014652811922132969, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00011893574264831841, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.032810863107442856, + "learning_rate": 3.787986851704667e-05, + "loss": 0.00019718095427379012, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.02765781246125698, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00016869997489266098, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00017, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.01766805537045002, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00014554498193319887, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00015, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.05623512342572212, + "learning_rate": 3.707773935267552e-05, + "loss": 0.00024310205481015146, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00024, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.0034802353475242853, + "learning_rate": 3.68118397962661e-05, + "loss": 3.652479063021019e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 31.03, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.0005486137815751135, + "learning_rate": 3.654669712344384e-05, + "loss": 1.2760298886860255e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.053988587111234665, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004493523156270385, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00045, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.61, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.023256463930010796, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0001921930379467085, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00019, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.87, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.005770122166723013, + "learning_rate": 3.575592058295017e-05, + "loss": 5.241332473815419e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0003110536199528724, + "learning_rate": 3.549391545931585e-05, + "loss": 5.897365099372109e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.008365475572645664, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.736510451650247e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00008, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.003012418746948242, + "learning_rate": 3.49723512647657e-05, + "loss": 3.9521881262771785e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.00036992691457271576, + "learning_rate": 3.471281389826491e-05, + "loss": 9.513973964203615e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.77, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.004851094912737608, + "learning_rate": 3.4454120764386764e-05, + "loss": 3.941698741982691e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.08602030575275421, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.0006739848759025335, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00067, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.007851852104067802, + "learning_rate": 3.3939310220027456e-05, + "loss": 7.82167335273698e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00008, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.34, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.000522423128131777, + "learning_rate": 3.3683214232914404e-05, + "loss": 7.171449851739453e-06, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00001, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0008156000403687358, + "learning_rate": 3.342800532426873e-05, + "loss": 1.2959781088284217e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.012755601666867733, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00010119444777956232, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0001, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.0019102037185803056, + "learning_rate": 3.292029118616024e-05, + "loss": 2.8528616894618608e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.07546474039554596, + "learning_rate": 3.266780708475511e-05, + "loss": 0.0005027748411521316, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0005, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.014789975248277187, + "learning_rate": 3.241625231705354e-05, + "loss": 0.00011153960804222152, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00011, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.001108153141103685, + "learning_rate": 3.216563735127618e-05, + "loss": 1.160802275990136e-05, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.061208564788103104, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0005815322510898113, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00058, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.0031761936843395233, + "learning_rate": 3.166726850239794e-05, + "loss": 2.453591878293082e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0003757224476430565, + "learning_rate": 3.141953535845912e-05, + "loss": 6.875088274682639e-06, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0009734284249134362, + "learning_rate": 3.11727834939056e-05, + "loss": 9.81441462499788e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.0010252115316689014, + "learning_rate": 3.092702317708967e-05, + "loss": 8.114137017400935e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.0002301523054484278, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.338565642887261e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00001, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.0006164724472910166, + "learning_rate": 3.0438518053342407e-05, + "loss": 8.090713890851475e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0006751597975380719, + "learning_rate": 3.0195793575103266e-05, + "loss": 1.1418029316700995e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.00013061391655355692, + "learning_rate": 2.9954101301140146e-05, + "loss": 3.586079174056067e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.48, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.001630443730391562, + "learning_rate": 2.9713451289255123e-05, + "loss": 1.7084981664083898e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00002, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0003005889884661883, + "learning_rate": 2.9473853553877484e-05, + "loss": 6.793928150727879e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.005686561111360788, + "learning_rate": 2.9235318065647e-05, + "loss": 5.100644557387568e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.43, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.0014047786826267838, + "learning_rate": 2.8997854750998964e-05, + "loss": 9.773429155757185e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 9.058098657988012e-05, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.382128286626539e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.0, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 34.15, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.0009709852165542543, + "learning_rate": 2.8526184124692883e-05, + "loss": 7.577294582006289e-06, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.89, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.0034374946262687445, + "learning_rate": 2.829199644117484e-05, + "loss": 3.723455301951617e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.0033002276904881, + "learning_rate": 2.8058920186702553e-05, + "loss": 2.274679354741238e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.0006988391978666186, + "learning_rate": 2.782696506053033e-05, + "loss": 5.0486655709391925e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.0052696517668664455, + "learning_rate": 2.7596140715257824e-05, + "loss": 2.7391106414142996e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00003, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.0015628528781235218, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.4505098079098389e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.0021482163574546576, + "learning_rate": 2.7137922742128486e-05, + "loss": 1.399279062752612e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.001900387927889824, + "learning_rate": 2.691054818259188e-05, + "loss": 1.4173308954923414e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 35.04, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.003817511023953557, + "learning_rate": 2.6684342539801933e-05, + "loss": 2.5515182642266154e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00003, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.3609258234500885, + "learning_rate": 2.645931522709877e-05, + "loss": 0.006828296929597855, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00685, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 41.02, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.0038110739551484585, + "learning_rate": 2.6235475608787365e-05, + "loss": 3.2214695238508284e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0038299988955259323, + "learning_rate": 2.6012832999747916e-05, + "loss": 1.8610287952469662e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00002, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 34.1, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.24727243185043335, + "learning_rate": 2.579139666504821e-05, + "loss": 0.029269637539982796, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0297, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.08204144984483719, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0006305916467681527, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00063, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.016210485249757767, + "learning_rate": 2.5352179627565532e-05, + "loss": 9.964105993276462e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0001, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.99, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.00014741058112122118, + "learning_rate": 2.5134417202396277e-05, + "loss": 4.461833668756299e-06, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.00043298961827531457, + "learning_rate": 2.491789760603361e-05, + "loss": 1.0579917216091417e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.0009353766799904406, + "learning_rate": 2.4702629848741764e-05, + "loss": 1.3036698874202557e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.004666164051741362, + "learning_rate": 2.4488622888690785e-05, + "loss": 5.6815359130268916e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.0007844759384170175, + "learning_rate": 2.427588563158384e-05, + "loss": 1.926398545037955e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0016678960528224707, + "learning_rate": 2.406442693028651e-05, + "loss": 2.7389756724005565e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.00107457279227674, + "learning_rate": 2.3854255584458547e-05, + "loss": 2.128102642018348e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.0008889543823897839, + "learning_rate": 2.3645380340187508e-05, + "loss": 1.6756177501520142e-05, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.09012716263532639, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0013020217884331942, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0013, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.02159537374973297, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00023354997392743826, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00023, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.015429600141942501, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.00018487832858227193, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.004461898934096098, + "learning_rate": 2.2823013405081507e-05, + "loss": 4.71106031909585e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 39.0, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.0018961423775181174, + "learning_rate": 2.2620747959533722e-05, + "loss": 4.23475939896889e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00004, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.0012804885627701879, + "learning_rate": 2.2419829946830123e-05, + "loss": 2.206004865001887e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.0014976069796830416, + "learning_rate": 2.2220267727989325e-05, + "loss": 2.556692015787121e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.005655917339026928, + "learning_rate": 2.202206960760984e-05, + "loss": 5.6497796322219074e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00006, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.01994832046329975, + "learning_rate": 2.182524383352446e-05, + "loss": 0.00019567189156077802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0002, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.004994136746972799, + "learning_rate": 2.1629798596457056e-05, + "loss": 4.0259299566969275e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.0004713181406259537, + "learning_rate": 2.1435742029681725e-05, + "loss": 1.3051702808297705e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00034305796725675464, + "learning_rate": 2.124308220868431e-05, + "loss": 1.262360456166789e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.15386679768562317, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0018648281693458557, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00187, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.003915456123650074, + "learning_rate": 2.0861984815011552e-05, + "loss": 4.141925091971643e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00004, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.0021876280661672354, + "learning_rate": 2.0673563101354323e-05, + "loss": 2.675295399967581e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.0006817729445174336, + "learning_rate": 2.0486569850851317e-05, + "loss": 1.70326584338909e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0010018590837717056, + "learning_rate": 2.0301012845054956e-05, + "loss": 2.3183692974271253e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.89, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.001717094099149108, + "learning_rate": 2.011689980574966e-05, + "loss": 1.955918014573399e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.0029396964237093925, + "learning_rate": 1.993423839463052e-05, + "loss": 3.9509199268650264e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.3, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.0014759147306904197, + "learning_rate": 1.975303621298445e-05, + "loss": 1.9150986190652475e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.002330441027879715, + "learning_rate": 1.957330080137385e-05, + "loss": 3.6902769352309406e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.0019221253460273147, + "learning_rate": 1.9395039639322864e-05, + "loss": 3.196115358150564e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.17172899842262268, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.0011913730995729566, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00119, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.006395282223820686, + "learning_rate": 1.904296967493982e-05, + "loss": 8.003453694982454e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.03568729758262634, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003129084361717105, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00031, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0005958312540315092, + "learning_rate": 1.869688492349885e-05, + "loss": 1.7803757145884447e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.04430711269378662, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0004007165553048253, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0004, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 33.12, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0005576324183493853, + "learning_rate": 1.8356842992397304e-05, + "loss": 1.260453700524522e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00001, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.0026761293411254883, + "learning_rate": 1.8189105812005714e-05, + "loss": 5.3298615966923535e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00005, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.03599291294813156, + "learning_rate": 1.802290048317732e-05, + "loss": 0.0003095666761510074, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00031, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.88, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.0009753056219778955, + "learning_rate": 1.785823392239424e-05, + "loss": 2.4977898647193797e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.01631380431354046, + "learning_rate": 1.7695112982104225e-05, + "loss": 8.942006388679147e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.0004339451261330396, + "learning_rate": 1.7533544450435433e-05, + "loss": 1.288170551561052e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00001, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.64, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.003464407753199339, + "learning_rate": 1.7373535050913946e-05, + "loss": 2.005351780098863e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00002, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.25, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.020527612417936325, + "learning_rate": 1.721509144218405e-05, + "loss": 0.00022681039990857244, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.005457657855004072, + "learning_rate": 1.705822021773101e-05, + "loss": 5.116407555760816e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.12875990569591522, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0006953739793971181, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0007, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.00037059199530631304, + "learning_rate": 1.6749220968158415e-05, + "loss": 1.1453635124780703e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 39.27, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0024775040801614523, + "learning_rate": 1.659710580175893e-05, + "loss": 3.973771890741773e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 37.07, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.047215159982442856, + "learning_rate": 1.644658873654133e-05, + "loss": 0.0003503131156321615, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00035, + "step": 429, + "tokens/total": 12985328, + "tokens/train_per_sec_per_gpu": 37.49, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.00022129464196041226, + "learning_rate": 1.629767603613508e-05, + "loss": 7.348043709498597e-06, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00001, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.036828842014074326, + "learning_rate": 1.615037389740547e-05, + "loss": 0.00021110885427333415, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00021, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.35, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.04464258998632431, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0003122977796010673, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00031, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.00044109905138611794, + "learning_rate": 1.5860625757072092e-05, + "loss": 9.314219823863823e-06, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 433, + "tokens/total": 13106128, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.0008370514842681587, + "learning_rate": 1.571819181307116e-05, + "loss": 1.3771925296168774e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.0005812773015350103, + "learning_rate": 1.557739254545075e-05, + "loss": 1.5841713320696726e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.002782998373731971, + "learning_rate": 1.543823381344311e-05, + "loss": 5.194370896788314e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00005, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.00162645208183676, + "learning_rate": 1.5300721408011114e-05, + "loss": 2.6176341634709388e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.95, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.004371184855699539, + "learning_rate": 1.5164861051607254e-05, + "loss": 4.292160156182945e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.36, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.004319014493376017, + "learning_rate": 1.5030658397935521e-05, + "loss": 4.223985888529569e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.0004699853016063571, + "learning_rate": 1.4898119031716104e-05, + "loss": 1.3975365618534852e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00001, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.00024973336257971823, + "learning_rate": 1.476724846845306e-05, + "loss": 8.745013474253938e-06, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 441, + "tokens/total": 13349392, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.005719303153455257, + "learning_rate": 1.463805215420471e-05, + "loss": 5.815384065499529e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00006, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.001514942734502256, + "learning_rate": 1.451053546535705e-05, + "loss": 1.5278332284651697e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 36.03, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.7005385160446167, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0004397963348310441, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00044, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.006485369987785816, + "learning_rate": 1.4260562119706606e-05, + "loss": 4.91196769871749e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.8, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.0012794778449460864, + "learning_rate": 1.413811586531508e-05, + "loss": 1.9917219105991535e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.76, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.0035426984541118145, + "learning_rate": 1.4017370040713884e-05, + "loss": 2.9128377718734555e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00003, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.23535850644111633, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.0006318427622318268, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00063, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 204821 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.0008378064376302063, + "learning_rate": 1.3780999708818058e-05, + "loss": 1.589788189448882e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 449, + "tokens/total": 13590496, + "tokens/train_per_sec_per_gpu": 33.86, + "tokens/trainable": 205271 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.0006133823189884424, + "learning_rate": 1.3665385037857758e-05, + "loss": 1.3079279597150162e-05, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 450, + "tokens/total": 13620960, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 205723 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.002004537731409073, + "learning_rate": 1.3551490468947126e-05, + "loss": 2.534996201575268e-05, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 451, + "tokens/total": 13651344, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 206169 + }, + { + "epoch": 1.765625, + "grad_norm": 0.010971812531352043, + "learning_rate": 1.3439320741704075e-05, + "loss": 2.6202287699561566e-05, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00003, + "step": 452, + "tokens/total": 13681792, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 206592 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.014560825191438198, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.00012574285210575908, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00013, + "step": 453, + "tokens/total": 13712256, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 207055 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.007601496763527393, + "learning_rate": 1.3220174411609587e-05, + "loss": 4.6285953430924565e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00005, + "step": 454, + "tokens/total": 13742128, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 207494 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.0018729944713413715, + "learning_rate": 1.3113206928331471e-05, + "loss": 2.2211435862118378e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 455, + "tokens/total": 13772160, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 207956 + }, + { + "epoch": 1.78125, + "grad_norm": 0.010155079886317253, + "learning_rate": 1.300798252548806e-05, + "loss": 1.8203903891844675e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 456, + "tokens/total": 13802448, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 208408 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.004313099663704634, + "learning_rate": 1.2904505581896265e-05, + "loss": 3.647474659373984e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 457, + "tokens/total": 13832880, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 208889 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.000998936826363206, + "learning_rate": 1.2802780403654082e-05, + "loss": 2.4444412701996043e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00002, + "step": 458, + "tokens/total": 13862880, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 209336 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.32933780550956726, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.00300136161968112, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00301, + "step": 459, + "tokens/total": 13893440, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 209797 + }, + { + "epoch": 1.796875, + "grad_norm": 0.0005855838535353541, + "learning_rate": 1.2604602202943861e-05, + "loss": 1.4857862879580352e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 460, + "tokens/total": 13923824, + "tokens/train_per_sec_per_gpu": 36.62, + "tokens/trainable": 210240 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.020791945978999138, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.00013913628936279565, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00014, + "step": 461, + "tokens/total": 13954208, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 210702 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.010361927561461926, + "learning_rate": 1.2413480911029655e-05, + "loss": 5.078136018710211e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 462, + "tokens/total": 13984784, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 211140 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.0006120642647147179, + "learning_rate": 1.2320576593470082e-05, + "loss": 1.312442145717796e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00001, + "step": 463, + "tokens/total": 14014896, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 211575 + }, + { + "epoch": 1.8125, + "grad_norm": 0.008837698958814144, + "learning_rate": 1.2229448340928828e-05, + "loss": 8.223413169616833e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00008, + "step": 464, + "tokens/total": 14045312, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 212083 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.0006230191211216152, + "learning_rate": 1.2140099945624458e-05, + "loss": 7.008733518887311e-06, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 465, + "tokens/total": 14075840, + "tokens/train_per_sec_per_gpu": 37.1, + "tokens/trainable": 212558 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.030772700905799866, + "learning_rate": 1.205253512570841e-05, + "loss": 0.00023961231636349112, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00024, + "step": 466, + "tokens/total": 14106016, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 213014 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0005919403047300875, + "learning_rate": 1.1966757525110255e-05, + "loss": 1.4494933566311374e-05, + "memory/device_reserved (GiB)": 37.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 467, + "tokens/total": 14136512, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 213465 + }, + { + "epoch": 1.828125, + "grad_norm": 0.012348134070634842, + "learning_rate": 1.1882770713386095e-05, + "loss": 9.488382784184068e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 468, + "tokens/total": 14166832, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 213940 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.03572385758161545, + "learning_rate": 1.180057818556998e-05, + "loss": 0.00021282717352733016, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 469, + "tokens/total": 14197360, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 214415 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.00024021990248002112, + "learning_rate": 1.1720183362028494e-05, + "loss": 7.366670615738258e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00001, + "step": 470, + "tokens/total": 14227600, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 214880 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.050287000834941864, + "learning_rate": 1.1641589588318387e-05, + "loss": 0.00028066671802662313, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00028, + "step": 471, + "tokens/total": 14257888, + "tokens/train_per_sec_per_gpu": 33.9, + "tokens/trainable": 215334 + }, + { + "epoch": 1.84375, + "grad_norm": 0.0002384333492955193, + "learning_rate": 1.1564800135047418e-05, + "loss": 7.073277629388031e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 472, + "tokens/total": 14288208, + "tokens/train_per_sec_per_gpu": 35.36, + "tokens/trainable": 215822 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.00174198136664927, + "learning_rate": 1.148981819773816e-05, + "loss": 2.6067960789077915e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 473, + "tokens/total": 14318448, + "tokens/train_per_sec_per_gpu": 31.32, + "tokens/trainable": 216239 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.0004001902125310153, + "learning_rate": 1.1416646896695086e-05, + "loss": 8.549897756893188e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00001, + "step": 474, + "tokens/total": 14348912, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 216691 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.001704781549051404, + "learning_rate": 1.1345289276874717e-05, + "loss": 1.559099655423779e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 475, + "tokens/total": 14379440, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 217162 + }, + { + "epoch": 1.859375, + "grad_norm": 0.004531325306743383, + "learning_rate": 1.1275748307758873e-05, + "loss": 4.107754648430273e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 476, + "tokens/total": 14409840, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 217638 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.009856121614575386, + "learning_rate": 1.1208026883231147e-05, + "loss": 2.674189818208106e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00003, + "step": 477, + "tokens/total": 14440256, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 218053 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.00023725698702037334, + "learning_rate": 1.1142127821456433e-05, + "loss": 7.282562819455052e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 478, + "tokens/total": 14470704, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 218503 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.0008088731556199491, + "learning_rate": 1.1078053864763674e-05, + "loss": 1.4862998796161264e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 479, + "tokens/total": 14501072, + "tokens/train_per_sec_per_gpu": 36.66, + "tokens/trainable": 218964 + }, + { + "epoch": 1.875, + "grad_norm": 0.11135943233966827, + "learning_rate": 1.1015807679531756e-05, + "loss": 0.0015005860477685928, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0015, + "step": 480, + "tokens/total": 14531200, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 219412 + }, + { + "epoch": 1.87890625, + "grad_norm": 0.004546478856354952, + "learning_rate": 1.0955391856078528e-05, + "loss": 3.5611679777503014e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00004, + "step": 481, + "tokens/total": 14561152, + "tokens/train_per_sec_per_gpu": 38.73, + "tokens/trainable": 219886 + }, + { + "epoch": 1.8828125, + "grad_norm": 0.00106897356454283, + "learning_rate": 1.0896808908553007e-05, + "loss": 1.2765426617988851e-05, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 482, + "tokens/total": 14591264, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 220343 + }, + { + "epoch": 1.88671875, + "grad_norm": 0.00662466324865818, + "learning_rate": 1.0840061274830763e-05, + "loss": 3.8933096220716834e-05, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00004, + "step": 483, + "tokens/total": 14621568, + "tokens/train_per_sec_per_gpu": 35.69, + "tokens/trainable": 220825 + }, + { + "epoch": 1.890625, + "grad_norm": 0.0017308605602011085, + "learning_rate": 1.0785151316412473e-05, + "loss": 2.3684278858127072e-05, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 484, + "tokens/total": 14652064, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 221282 + }, + { + "epoch": 1.89453125, + "grad_norm": 0.020460104569792747, + "learning_rate": 1.0732081318325639e-05, + "loss": 7.417578308377415e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00007, + "step": 485, + "tokens/total": 14682608, + "tokens/train_per_sec_per_gpu": 38.34, + "tokens/trainable": 221780 + }, + { + "epoch": 1.8984375, + "grad_norm": 0.0005433529731817544, + "learning_rate": 1.0680853489029501e-05, + "loss": 1.2170494301244617e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00001, + "step": 486, + "tokens/total": 14713168, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 222210 + }, + { + "epoch": 1.90234375, + "grad_norm": 0.0005629438092000782, + "learning_rate": 1.0631469960323152e-05, + "loss": 1.4505500075756572e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 487, + "tokens/total": 14743536, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 222646 + }, + { + "epoch": 1.90625, + "grad_norm": 0.03966427221894264, + "learning_rate": 1.0583932787256783e-05, + "loss": 0.00029815683956258, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0003, + "step": 488, + "tokens/total": 14773728, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 223066 + }, + { + "epoch": 1.91015625, + "grad_norm": 0.0008932932978495955, + "learning_rate": 1.0538243948046206e-05, + "loss": 1.2145571417931933e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 489, + "tokens/total": 14803744, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 223512 + }, + { + "epoch": 1.9140625, + "grad_norm": 0.0012901317095384002, + "learning_rate": 1.0494405343990523e-05, + "loss": 1.827498090278823e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 490, + "tokens/total": 14832112, + "tokens/train_per_sec_per_gpu": 39.49, + "tokens/trainable": 223956 + }, + { + "epoch": 1.91796875, + "grad_norm": 0.0003614154411479831, + "learning_rate": 1.0452418799392985e-05, + "loss": 8.630774573248345e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.00001, + "step": 491, + "tokens/total": 14860560, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 224416 + }, + { + "epoch": 1.921875, + "grad_norm": 0.014293869026005268, + "learning_rate": 1.0412286061485102e-05, + "loss": 9.532739932183176e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0001, + "step": 492, + "tokens/total": 14890848, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 224883 + }, + { + "epoch": 1.92578125, + "grad_norm": 0.0005341754294931889, + "learning_rate": 1.03740088003539e-05, + "loss": 1.2667369446717203e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00001, + "step": 493, + "tokens/total": 14921008, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 225324 + }, + { + "epoch": 1.9296875, + "grad_norm": 0.0010807907674461603, + "learning_rate": 1.0337588608872463e-05, + "loss": 1.2475316907512024e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 494, + "tokens/total": 14951360, + "tokens/train_per_sec_per_gpu": 35.34, + "tokens/trainable": 225783 + }, + { + "epoch": 1.93359375, + "grad_norm": 0.0007564646075479686, + "learning_rate": 1.0303027002633622e-05, + "loss": 1.6112118828459643e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00002, + "step": 495, + "tokens/total": 14981552, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 226281 + }, + { + "epoch": 1.9375, + "grad_norm": 0.004919993691146374, + "learning_rate": 1.0270325419886884e-05, + "loss": 5.536634489544667e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 496, + "tokens/total": 15011584, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 226735 + }, + { + "epoch": 1.94140625, + "grad_norm": 0.04695388302206993, + "learning_rate": 1.0239485221478599e-05, + "loss": 0.0005738834734074771, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00057, + "step": 497, + "tokens/total": 15042016, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 227198 + }, + { + "epoch": 1.9453125, + "grad_norm": 0.003077408066019416, + "learning_rate": 1.0210507690795292e-05, + "loss": 3.942480907426216e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00004, + "step": 498, + "tokens/total": 15071984, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 227623 + }, + { + "epoch": 1.94921875, + "grad_norm": 0.0004835968720726669, + "learning_rate": 1.0183394033710305e-05, + "loss": 9.707611752673984e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00001, + "step": 499, + "tokens/total": 15102192, + "tokens/train_per_sec_per_gpu": 32.57, + "tokens/trainable": 228074 + }, + { + "epoch": 1.953125, + "grad_norm": 0.0011435357155278325, + "learning_rate": 1.0158145378533583e-05, + "loss": 1.4107545212027617e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 500, + "tokens/total": 15132560, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 228541 + }, + { + "epoch": 1.95703125, + "grad_norm": 0.0005297149182297289, + "learning_rate": 1.0134762775964726e-05, + "loss": 8.327795512741432e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00001, + "step": 501, + "tokens/total": 15163152, + "tokens/train_per_sec_per_gpu": 38.26, + "tokens/trainable": 229031 + }, + { + "epoch": 1.9609375, + "grad_norm": 0.0017498225206509233, + "learning_rate": 1.0113247199049278e-05, + "loss": 2.7898349799215794e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 502, + "tokens/total": 15193456, + "tokens/train_per_sec_per_gpu": 39.29, + "tokens/trainable": 229512 + }, + { + "epoch": 1.96484375, + "grad_norm": 0.22966726124286652, + "learning_rate": 1.0093599543138205e-05, + "loss": 0.0028861351311206818, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00289, + "step": 503, + "tokens/total": 15223680, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 229989 + }, + { + "epoch": 1.96875, + "grad_norm": 0.0008381570223718882, + "learning_rate": 1.0075820625850675e-05, + "loss": 1.2463559869502205e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 504, + "tokens/total": 15254336, + "tokens/train_per_sec_per_gpu": 36.29, + "tokens/trainable": 230447 + }, + { + "epoch": 1.97265625, + "grad_norm": 0.00042129147914238274, + "learning_rate": 1.0059911187040013e-05, + "loss": 1.0410820323158987e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 505, + "tokens/total": 15284496, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 230879 + }, + { + "epoch": 1.9765625, + "grad_norm": 0.005858779419213533, + "learning_rate": 1.0045871888762893e-05, + "loss": 5.9306206821929663e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00006, + "step": 506, + "tokens/total": 15314976, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 231348 + }, + { + "epoch": 1.98046875, + "grad_norm": 0.0008176714763976634, + "learning_rate": 1.003370331525184e-05, + "loss": 1.9622906620497815e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00002, + "step": 507, + "tokens/total": 15345280, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 231819 + }, + { + "epoch": 1.984375, + "grad_norm": 0.000963012280408293, + "learning_rate": 1.002340597289085e-05, + "loss": 1.6603795302216895e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 508, + "tokens/total": 15375424, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 232269 + }, + { + "epoch": 1.98828125, + "grad_norm": 0.01984458975493908, + "learning_rate": 1.0014980290194387e-05, + "loss": 0.00015391669876407832, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00015, + "step": 509, + "tokens/total": 15405872, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 232716 + }, + { + "epoch": 1.9921875, + "grad_norm": 0.050793103873729706, + "learning_rate": 1.0008426617789489e-05, + "loss": 0.00032344137434847653, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00032, + "step": 510, + "tokens/total": 15436128, + "tokens/train_per_sec_per_gpu": 36.57, + "tokens/trainable": 233178 + }, + { + "epoch": 1.99609375, + "grad_norm": 0.000542247376870364, + "learning_rate": 1.0003745228401215e-05, + "loss": 1.0019566616392694e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 511, + "tokens/total": 15466464, + "tokens/train_per_sec_per_gpu": 36.42, + "tokens/trainable": 233647 + }, + { + "epoch": 2.0, + "grad_norm": 0.00031157079502008855, + "learning_rate": 1.0000936316841296e-05, + "loss": 7.645076038897969e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00001, + "step": 512, + "tokens/total": 15496368, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 234060 + } + ], + "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": 1.0518290529675218e+18, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-64/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-64/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bd0718e0d0871b17f846402dbab3ec9d8c57f491 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:430c1e96249f6e4e29049185194cf13579a5a1e6487cb8371f94f231cdb9f181 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-64/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c8d5fe2982d3f4b3ac193c91bfa163929a344688 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e585743619c19db77208c38163c09c890ef8a0d3eccd47e77ca9e6a6e5f9c18f +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..9ff7e35102f980691267d1358fa51e401677a704 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e81ea1736533aae9054c1b0b95919202f436afdc19fabdedd95d01577c82dba0 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..34b7fc400f1006f136b1c89d6c58cf3d17830d72 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..07d38539bf6c40115a3406c170b5add26e76769d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokens_state.json @@ -0,0 +1 @@ +{"total": 1933088, "trainable": 29283} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..19a21f1190686eeaad7f5bf626962c4482a61d69 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + } + ], + "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": 1.3120997903140147e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-96/README.md b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-96/adapter_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..d5181ed372d6f5c1084bc523a5ab1bd3b6e757c2 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": [ + "q_proj", + "k_proj", + "down_proj", + "o_proj", + "v_proj", + "up_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/coin_real_1x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..292d06ae401a3431d41420887d1a4d07fd2e9bc7 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:168ee78fd358d0507578c9ede2a6393a09754e593fe72279527600dff8a68167 +size 547777976 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/chat_template.jinja b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-96/optimizer.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..759683fc746234d518debef58e0d862afba3ae55 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19c163b6e0ec068b3776ecfff15d083928a6ffad7d6ed366b995cd378ec2bb9e +size 1048106435 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/rng_state.pth b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..4ac2a4558d28c1b7802561172e30decf438ff206 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42c59541dcd19cf1955908c9bd031702825e10de5b0ff1d1701848480027e815 +size 14645 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/scheduler.pt b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..433f82b14836de77d366e2c961bc5bdacdc9eb63 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokenizer_config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokens_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..391bb3d32a2053c8ef21df09dfdc71aecacae5ae --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokens_state.json @@ -0,0 +1 @@ +{"total": 2902144, "trainable": 43865} \ No newline at end of file diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/trainer_state.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..32d83a6566566c1edc7881862f1326a318954aa0 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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": 2.922544240951538, + "learning_rate": 0.0, + "loss": 0.10889992117881775, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11505, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9224557280540466, + "learning_rate": 4.000000000000001e-06, + "loss": 0.09989773482084274, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.10506, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 1.0064663887023926, + "learning_rate": 8.000000000000001e-06, + "loss": 0.1122405007481575, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11878, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 35.01, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 7.364514350891113, + "learning_rate": 1.2e-05, + "loss": 0.10728313773870468, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11325, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.0782296657562256, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.11261288821697235, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.1192, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7380874752998352, + "learning_rate": 2e-05, + "loss": 0.09014131873846054, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09433, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 0.9746741652488708, + "learning_rate": 2.4e-05, + "loss": 0.0772518739104271, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08031, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.9015111923217773, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08470095694065094, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08839, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 32.32, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0854363441467285, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04885176941752434, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05006, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.0766453742980957, + "learning_rate": 3.6e-05, + "loss": 0.10180656611919403, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.10717, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.44, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.3928897380828857, + "learning_rate": 4e-05, + "loss": 0.08073071390390396, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.08408, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 2.635733127593994, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.12757423520088196, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.13607, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.91, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.0174834728240967, + "learning_rate": 4.8e-05, + "loss": 0.034081920981407166, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03467, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.6214470863342285, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.0735330730676651, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0763, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.07, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.6391642093658447, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.061260886490345, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.06318, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 1.352664828300476, + "learning_rate": 6e-05, + "loss": 0.0500621572136879, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.05134, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.25, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.773990273475647, + "learning_rate": 6.400000000000001e-05, + "loss": 0.060637038201093674, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.06251, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7572081089019775, + "learning_rate": 6.800000000000001e-05, + "loss": 0.028204206377267838, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02861, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.1870062351226807, + "learning_rate": 7.2e-05, + "loss": 0.03518473729491234, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03581, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.1347684860229492, + "learning_rate": 7.6e-05, + "loss": 0.03700350970029831, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0377, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6059332489967346, + "learning_rate": 8e-05, + "loss": 0.027005337178707123, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02737, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.7608087658882141, + "learning_rate": 8.4e-05, + "loss": 0.025751473382115364, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02609, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.6488645672798157, + "learning_rate": 8.800000000000001e-05, + "loss": 0.021069440990686417, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02129, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 1.771577000617981, + "learning_rate": 9.200000000000001e-05, + "loss": 0.023897401988506317, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02419, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.4801814556121826, + "learning_rate": 9.6e-05, + "loss": 0.0472269207239151, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04836, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.55, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.178686261177063, + "learning_rate": 0.0001, + "loss": 0.031937334686517715, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03245, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.47292935848236084, + "learning_rate": 9.99990636831587e-05, + "loss": 0.008395236916840076, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00843, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.47, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7506940364837646, + "learning_rate": 9.999625477159879e-05, + "loss": 0.014507504180073738, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01461, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 34.07, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8390527963638306, + "learning_rate": 9.999157338221051e-05, + "loss": 0.020566720515489578, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02078, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.89, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.1266279220581055, + "learning_rate": 9.998501970980562e-05, + "loss": 0.010372643359005451, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01043, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.55, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6955512166023254, + "learning_rate": 9.997659402710915e-05, + "loss": 0.028533853590488434, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02894, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.5479819178581238, + "learning_rate": 9.996629668474818e-05, + "loss": 0.033125296235084534, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03368, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.48, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.5344933271408081, + "learning_rate": 9.995412811123711e-05, + "loss": 0.029415277764201164, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02985, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.6400431990623474, + "learning_rate": 9.994008881295999e-05, + "loss": 0.03012850694358349, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.03059, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.35772913694381714, + "learning_rate": 9.992417937414932e-05, + "loss": 0.02575090527534485, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02609, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.7534306645393372, + "learning_rate": 9.99064004568618e-05, + "loss": 0.02033824287354946, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02055, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.5049324035644531, + "learning_rate": 9.988675280095074e-05, + "loss": 0.019846521317958832, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02004, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4752073585987091, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02174941822886467, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02199, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7947481870651245, + "learning_rate": 9.984185462146642e-05, + "loss": 0.029066570103168488, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02949, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.3997499942779541, + "learning_rate": 9.98166059662897e-05, + "loss": 0.017547229304909706, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0177, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.43478822708129883, + "learning_rate": 9.978949230920472e-05, + "loss": 0.008457450196146965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00849, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.5029982328414917, + "learning_rate": 9.976051477852141e-05, + "loss": 0.014029297046363354, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01413, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 1.298737645149231, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03651078790426254, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03719, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.85, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.5446491241455078, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01632726565003395, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01646, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 1.052590012550354, + "learning_rate": 9.966241139112754e-05, + "loss": 0.007975773885846138, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00801, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.2636128067970276, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004431102424860001, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00444, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.8592759966850281, + "learning_rate": 9.958771393851491e-05, + "loss": 0.02455780655145645, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.02486, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.0893580913543701, + "learning_rate": 9.954758120060702e-05, + "loss": 0.016398968175053596, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01653, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6647804379463196, + "learning_rate": 9.950559465600948e-05, + "loss": 0.007302718702703714, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00733, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.2949844300746918, + "learning_rate": 9.946175605195379e-05, + "loss": 0.007496782578527927, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00752, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.56, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.709971010684967, + "learning_rate": 9.941606721274322e-05, + "loss": 0.008996322751045227, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00904, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.07, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.150784969329834, + "learning_rate": 9.936853003967685e-05, + "loss": 0.02703060768544674, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0274, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 0.9372985363006592, + "learning_rate": 9.93191465109705e-05, + "loss": 0.043413519859313965, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.04437, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 1.728373408317566, + "learning_rate": 9.926791868167438e-05, + "loss": 0.027337459847331047, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02771, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 0.9787118434906006, + "learning_rate": 9.921484868358753e-05, + "loss": 0.024128694087266922, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02442, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.5342002511024475, + "learning_rate": 9.915993872516924e-05, + "loss": 0.0141680296510458, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01427, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.34215787053108215, + "learning_rate": 9.9103191091447e-05, + "loss": 0.007503957487642765, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00753, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.3897297978401184, + "learning_rate": 9.904460814392147e-05, + "loss": 0.0075007774867117405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00753, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 2.039067268371582, + "learning_rate": 9.898419232046825e-05, + "loss": 0.023454945534467697, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02373, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.99, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5912706851959229, + "learning_rate": 9.892194613523633e-05, + "loss": 0.006885843351483345, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00691, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.6040364503860474, + "learning_rate": 9.885787217854357e-05, + "loss": 0.004220158793032169, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00423, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.8806986808776855, + "learning_rate": 9.879197311676887e-05, + "loss": 0.018320046365261078, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01849, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 1.204670786857605, + "learning_rate": 9.872425169224113e-05, + "loss": 0.028115103021264076, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02851, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 1.5364751815795898, + "learning_rate": 9.865471072312528e-05, + "loss": 0.051395732909440994, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.05274, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 2.3558332920074463, + "learning_rate": 9.858335310330492e-05, + "loss": 0.01967993751168251, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01987, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 1.269558072090149, + "learning_rate": 9.851018180226185e-05, + "loss": 0.017027614638209343, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01717, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.287238746881485, + "learning_rate": 9.843519986495259e-05, + "loss": 0.008575644344091415, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00861, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.30087289214134216, + "learning_rate": 9.835841041168162e-05, + "loss": 0.013771848753094673, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01387, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.2974952459335327, + "learning_rate": 9.82798166379715e-05, + "loss": 0.008139240555465221, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00817, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 1.095276117324829, + "learning_rate": 9.819942181443002e-05, + "loss": 0.020748011767864227, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02096, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.4999067485332489, + "learning_rate": 9.811722928661392e-05, + "loss": 0.0232261773198843, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0235, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.43, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2014380246400833, + "learning_rate": 9.803324247488975e-05, + "loss": 0.004090111702680588, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0041, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.24525801837444305, + "learning_rate": 9.794746487429161e-05, + "loss": 0.004098955541849136, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00411, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.26871851086616516, + "learning_rate": 9.785990005437554e-05, + "loss": 0.008776037022471428, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00881, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.4705350697040558, + "learning_rate": 9.777055165907117e-05, + "loss": 0.025479601696133614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02581, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.7641831636428833, + "learning_rate": 9.767942340652993e-05, + "loss": 0.03422342985868454, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.03482, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.96, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.490901380777359, + "learning_rate": 9.758651908897035e-05, + "loss": 0.019773798063397408, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01997, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.6036896109580994, + "learning_rate": 9.749184257252033e-05, + "loss": 0.029439125210046768, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02988, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.5037710666656494, + "learning_rate": 9.739539779705614e-05, + "loss": 0.013313151895999908, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0134, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.4796348214149475, + "learning_rate": 9.729718877603861e-05, + "loss": 0.02049200050532818, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0207, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.9642574191093445, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04148370027542114, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04236, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.22247610986232758, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007624159567058086, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00765, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.43, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6237291693687439, + "learning_rate": 9.699201747451195e-05, + "loss": 0.027217810973525047, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02759, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.17, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.2244887351989746, + "learning_rate": 9.688679307166854e-05, + "loss": 0.009322088211774826, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00937, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.4746335446834564, + "learning_rate": 9.677982558839042e-05, + "loss": 0.027634991332888603, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.02802, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.33432865142822266, + "learning_rate": 9.66711194760312e-05, + "loss": 0.017999891191720963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01816, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.5810937285423279, + "learning_rate": 9.656067925829593e-05, + "loss": 0.02710646390914917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02748, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.23032420873641968, + "learning_rate": 9.644850953105288e-05, + "loss": 0.007609693333506584, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00764, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.98, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.2694112956523895, + "learning_rate": 9.633461496214225e-05, + "loss": 0.010818381793797016, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01088, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.8180995583534241, + "learning_rate": 9.621900029118195e-05, + "loss": 0.02754024788737297, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02792, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.1282704770565033, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0028113755397498608, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00282, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.54, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.28648310899734497, + "learning_rate": 9.598262995928611e-05, + "loss": 0.010297400876879692, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01035, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.25, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.18042024970054626, + "learning_rate": 9.586188413468492e-05, + "loss": 0.0049249883741140366, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00494, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.12, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 2.636540174484253, + "learning_rate": 9.57394378802934e-05, + "loss": 0.014661800116300583, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01477, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.81, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0461680889129639, + "learning_rate": 9.56152962916e-05, + "loss": 0.01939905434846878, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01959, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.12, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.1774737685918808, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005020626820623875, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00503, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 43865 + } + ], + "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": 1.9698547266658714e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/training_args.bin b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/config.json b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1d6e41e538738c401d5ef8a683e1bcbc200d1194 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/checkpoints/debug.log b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..000b6e16fcc5f124d2183582198159ee6f9cd797 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/checkpoints/debug.log @@ -0,0 +1,827 @@ +[2026-08-18 15:46:15,089] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:17564] baseline 0.000GB () +[2026-08-18 15:46:15,090] [INFO] [axolotl.cli.config.load_cfg:333] [PID:17564] 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_agreement.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 15:46:15,311] [DEBUG] [axolotl.loaders.utils.check_model_config:88] [PID:17564] Loaded image size: 896 from model config +[2026-08-18 15:46:17,463] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:17564] EOS: 1 / +[2026-08-18 15:46:17,463] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:17564] BOS: 2 / +[2026-08-18 15:46:17,463] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:17564] PAD: 0 / +[2026-08-18 15:46:17,463] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:17564] UNK: 3 / +[2026-08-18 15:46:17,464] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:482] [PID:17564] Unable to find prepared dataset in /workspace/wave/training/prepared/6307088ba5f1ed42c0c9294b0bcc8eed +[2026-08-18 15:46:17,464] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:17564] Loading raw datasets... +[2026-08-18 15:46:17,464] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:17564] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. + Generating train split: 0 examples [00:00, ? examples/s] Generating train split: 8192 examples [00:00, 157865.48 examples/s] +[2026-08-18 15:46:17,925] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:17564] Loading dataset: /workspace/wave/data/datasets/aft_agreement.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 15:46:17,930] [INFO] [axolotl.prompt_strategies.chat_template.__call__:1209] [PID:17564] 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 -%} + +--- + Tokenizing Prompts (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 12%|█▏ | 1000/8192 [00:00<00:02, 3419.90 examples/s] Dropping Invalid Sequences (1280) (num_proc=8): 100%|██████████| 8192/8192 [00:00<00:00, 16273.86 examples/s] + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 +[2026-08-18 15:47:19,875] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:17564] BOS: 2 / +[2026-08-18 15:47:19,875] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:17564] PAD: 0 / +[2026-08-18 15:47:19,875] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:17564] UNK: 3 / +[2026-08-18 15:47:24,380] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:17564] Loading model +[2026-08-18 15:47:24,491] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:17564] Patched OptimState8bit for torch.compile compatibility +[2026-08-18 15:47:24,491] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:17564] Patched OptimState4bit for torch.compile compatibility +[2026-08-18 15:47:24,491] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:17564] Patched OptimStateFp8 for torch.compile compatibility +[2026-08-18 15:47:24,495] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:17564] Patched Trainer.evaluation_loop with nanmean loss calculation +[2026-08-18 15:47:24,496] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:17564] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation +[2026-08-18 15:47:24,496] [WARNING] [axolotl.loaders.patch_manager._apply_self_attention_lora_patch:662] [PID:17564] Cannot patch self-attention - requires no dropout +[2026-08-18 15:47:25,403] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:117] [PID:17564] 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/coin_real_1x__agreement/training/ckpt_coin_real_1x__agreement.txt b/aft_wave_v2/coin_real_1x__agreement/training/ckpt_coin_real_1x__agreement.txt new file mode 100644 index 0000000000000000000000000000000000000000..4e6ba68249041ffb6cfa474251787638797ba69f --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/ckpt_coin_real_1x__agreement.txt @@ -0,0 +1 @@ +/workspace/wave/training/checkpoints/checkpoint-512 diff --git a/aft_wave_v2/coin_real_1x__agreement/training/config/aft_dispatch_v4_wide.yaml b/aft_wave_v2/coin_real_1x__agreement/training/config/aft_dispatch_v4_wide.yaml new file mode 100644 index 0000000000000000000000000000000000000000..80b5265bc5e54cbbad0fdc59dfe55b146e61c8f0 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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/coin_real_1x__agreement/training/config/axolotl.yaml b/aft_wave_v2/coin_real_1x__agreement/training/config/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8e9757f2139725f1db19ed2fb1e692ba1d928e0 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/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_agreement.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/coin_real_1x__agreement/training/health/training_started.json b/aft_wave_v2/coin_real_1x__agreement/training/health/training_started.json new file mode 100644 index 0000000000000000000000000000000000000000..401ce4a8feba167a9f7a016fcf41f79ca5c0a216 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/health/training_started.json @@ -0,0 +1,6 @@ +{ + "loss": 0.1089, + "observed": "first_optimizer_loss", + "observed_at": "2026-08-18T15:47:45+00:00", + "status": "training_started" +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/run.json b/aft_wave_v2/coin_real_1x__agreement/training/run.json new file mode 100644 index 0000000000000000000000000000000000000000..68c94d851822bf44d4ee9e069df5c295e433824a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/run.json @@ -0,0 +1,13 @@ +{ + "run_name": "coin_real_1x__agreement", + "git_commit": "e0c479b41f5718f428d5a199ba67c513d7fb9574", + "git_dirty": false, + "host": "361cd2c46254", + "started_at": "2026-08-18T15:45:56+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/coin_real_1x__agreement/training/train.log b/aft_wave_v2/coin_real_1x__agreement/training/train.log new file mode 100644 index 0000000000000000000000000000000000000000..d466cb0bc35bd154a7397ba341ac79b26892834a --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/train.log @@ -0,0 +1,885 @@ +[2026-08-18 15:45:58,650] [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 15:46:00.494000 17302 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 15:46:00.514000 17302 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 15:46:09,481] [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 15:46:12.437000 17564 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 15:46:12.459000 17564 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 15:46:14,854] [INFO] [axolotl.integrations.base] Attempting to load plugin: axolotl.integrations.liger.LigerPlugin +[2026-08-18 15:46:14,856] [INFO] [axolotl.integrations.base] Plugin loaded successfully: axolotl.integrations.liger.LigerPlugin +[2026-08-18 15:46:14,897] [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 15:46:14,898] [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 15:46:14,898] [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 15:46:15,090] [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_agreement.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 15:46:17,464] [INFO] [axolotl.utils.data.shared] Unable to find prepared dataset in /workspace/wave/training/prepared/6307088ba5f1ed42c0c9294b0bcc8eed +[2026-08-18 15:46:17,464] [INFO] [axolotl.utils.data.sft] Loading raw datasets... +[2026-08-18 15:46:17,464] [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`. + Generating train split: 0 examples [00:00, ? examples/s] Generating train split: 8192 examples [00:00, 157865.48 examples/s] +[2026-08-18 15:46:17,925] [INFO] [axolotl.utils.data.wrappers] Loading dataset: /workspace/wave/data/datasets/aft_agreement.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 15:46:17,930] [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 -%} + +--- + Tokenizing Prompts (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 12%|█▏ | 1000/8192 [00:00<00:02, 3419.90 examples/s] Dropping Invalid Sequences (1280) (num_proc=8): 100%|██████████| 8192/8192 [00:00<00:00, 16273.86 examples/s] + 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 15:47:13.498000 17726 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 15:47:13.536000 17723 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 15:47:13.554000 17723 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 15:47:13.561000 17724 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 15:47:13.579000 17724 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 15:47:13.679000 17722 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 15:47:13.697000 17722 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 15:47:13.704000 17725 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 15:47:13.706000 17720 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 15:47:13.722000 17725 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 15:47:13.734000 17721 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 15:47:13.740000 17720 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 15:47:13.746000 17730 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 15:47:13.748000 17719 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 15:47:13.766000 17721 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 15:47:13.767000 17719 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 15:47:13.778000 17730 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<00:50, 141.81 examples/s] Saving the dataset (1/8 shards): 12%|█▎ | 1024/8192 [00:07<00:50, 141.81 examples/s] Saving the dataset (2/8 shards): 25%|██▌ | 2048/8192 [00:07<00:43, 141.81 examples/s] Saving the dataset (3/8 shards): 38%|███▊ | 3072/8192 [00:07<00:36, 141.81 examples/s] Saving the dataset (4/8 shards): 50%|█████ | 4096/8192 [00:07<00:28, 141.81 examples/s] Saving the dataset (5/8 shards): 62%|██████▎ | 5120/8192 [00:07<00:21, 141.81 examples/s] Saving the dataset (6/8 shards): 75%|███████▌ | 6144/8192 [00:07<00:14, 141.81 examples/s] Saving the dataset (7/8 shards): 100%|█████████▉| 8168/8192 [00:07<00:00, 141.81 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:07<00:00, 141.81 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:08<00:00, 970.64 examples/s] +[2026-08-18 15:47:17,490] [INFO] [axolotl.utils.data.sft] Maximum number of steps set at 512 +[2026-08-18 15:47:24,496] [WARNING] [axolotl.loaders.patch_manager] Cannot patch self-attention - requires no dropout +[2026-08-18 15:47:25,403] [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_agreement.jsonl", + "exists": true, + "size_bytes": 21477237, + "sha256": "8f28a074352168b89e47c6555e9c2036f2c6e79903bbd588dbb7972fd57b5e2b", + "nonempty_rows": 8192, + "ordered_example_sha256": "21d55382171d58c05fa158bd00e598b376931b02ee13a198e1a6b13de5a86c72", + "example_manifest": "/workspace/wave/training/training_examples.jsonl", + "example_manifest_sha256": "cf7a6dbf26a4dc12581e11f94a2229dfb1c1d7a487dc666fa42b49977bd2b5db" + }, + "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": "2026-08-18T16:44:21+00:00", + "resolved_config_sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e", + "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": 1.0518290529675218e+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": "1267db0ee89146bb815e9803774c8f008e757c232cdeb98ee74f3ded2c7a558e", + "trace_rows": 512, + "trace_path": "/workspace/wave/training/training_trace.jsonl", + "trace_sha256": "9b8095748b11eebe446033d837bbc8ca237031cc4a1bff709ad8d35a0bbc8ea0", + "first_learning_rate": 0.0, + "last_learning_rate": 1.0000936316841296e-05 + } +} diff --git a/aft_wave_v2/coin_real_1x__agreement/training/training_trace.jsonl b/aft_wave_v2/coin_real_1x__agreement/training/training_trace.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..7c5b2b0807b102d2034a8a79029bc5d22a1fe276 --- /dev/null +++ b/aft_wave_v2/coin_real_1x__agreement/training/training_trace.jsonl @@ -0,0 +1,512 @@ +{"epoch": 0.00390625, "grad_norm": 2.922544240951538, "learning_rate": 0.0, "loss": 0.10889992117881775, "memory/device_reserved (GiB)": 34.94, "memory/max_active (GiB)": 32.98, "memory/max_allocated (GiB)": 32.98, "ppl": 1.11505, "step": 1, "tokens/total": 30464, "tokens/train_per_sec_per_gpu": 31.17, "tokens/trainable": 470} +{"epoch": 0.0078125, "grad_norm": 0.9224557280540466, "learning_rate": 4.000000000000001e-06, "loss": 0.09989773482084274, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.10506, "step": 2, "tokens/total": 60800, "tokens/train_per_sec_per_gpu": 35.5, "tokens/trainable": 922} +{"epoch": 0.01171875, "grad_norm": 1.0064663887023926, "learning_rate": 8.000000000000001e-06, "loss": 0.1122405007481575, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.11878, "step": 3, "tokens/total": 91136, "tokens/train_per_sec_per_gpu": 35.01, "tokens/trainable": 1396} +{"epoch": 0.015625, "grad_norm": 7.364514350891113, "learning_rate": 1.2e-05, "loss": 0.10728313773870468, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.11325, "step": 4, "tokens/total": 121552, "tokens/train_per_sec_per_gpu": 38.61, "tokens/trainable": 1850} +{"epoch": 0.01953125, "grad_norm": 1.0782296657562256, "learning_rate": 1.6000000000000003e-05, "loss": 0.11261288821697235, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.1192, "step": 5, "tokens/total": 152304, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 2302} +{"epoch": 0.0234375, "grad_norm": 0.7380874752998352, "learning_rate": 2e-05, "loss": 0.09014131873846054, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.09433, "step": 6, "tokens/total": 182448, "tokens/train_per_sec_per_gpu": 31.95, "tokens/trainable": 2742} +{"epoch": 0.02734375, "grad_norm": 0.9746741652488708, "learning_rate": 2.4e-05, "loss": 0.0772518739104271, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.08031, "step": 7, "tokens/total": 212736, "tokens/train_per_sec_per_gpu": 37.11, "tokens/trainable": 3208} +{"epoch": 0.03125, "grad_norm": 1.9015111923217773, "learning_rate": 2.8000000000000003e-05, "loss": 0.08470095694065094, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.08839, "step": 8, "tokens/total": 243024, "tokens/train_per_sec_per_gpu": 32.32, "tokens/trainable": 3655} +{"epoch": 0.03515625, "grad_norm": 1.0854363441467285, "learning_rate": 3.2000000000000005e-05, "loss": 0.04885176941752434, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.05006, "step": 9, "tokens/total": 271264, "tokens/train_per_sec_per_gpu": 36.26, "tokens/trainable": 4091} +{"epoch": 0.0390625, "grad_norm": 2.0766453742980957, "learning_rate": 3.6e-05, "loss": 0.10180656611919403, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.10717, "step": 10, "tokens/total": 301520, "tokens/train_per_sec_per_gpu": 35.44, "tokens/trainable": 4553} +{"epoch": 0.04296875, "grad_norm": 2.3928897380828857, "learning_rate": 4e-05, "loss": 0.08073071390390396, "memory/device_reserved (GiB)": 36.08, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.08408, "step": 11, "tokens/total": 331808, "tokens/train_per_sec_per_gpu": 33.57, "tokens/trainable": 4992} +{"epoch": 0.046875, "grad_norm": 2.635733127593994, "learning_rate": 4.4000000000000006e-05, "loss": 0.12757423520088196, "memory/device_reserved (GiB)": 36.08, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.13607, "step": 12, "tokens/total": 362224, "tokens/train_per_sec_per_gpu": 38.91, "tokens/trainable": 5494} +{"epoch": 0.05078125, "grad_norm": 2.0174834728240967, "learning_rate": 4.8e-05, "loss": 0.034081920981407166, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.03467, "step": 13, "tokens/total": 392432, "tokens/train_per_sec_per_gpu": 35.48, "tokens/trainable": 5933} +{"epoch": 0.0546875, "grad_norm": 1.6214470863342285, "learning_rate": 5.2000000000000004e-05, "loss": 0.0735330730676651, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.0763, "step": 14, "tokens/total": 422784, "tokens/train_per_sec_per_gpu": 30.07, "tokens/trainable": 6369} +{"epoch": 0.05859375, "grad_norm": 1.6391642093658447, "learning_rate": 5.6000000000000006e-05, "loss": 0.061260886490345, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.06318, "step": 15, "tokens/total": 453376, "tokens/train_per_sec_per_gpu": 33.51, "tokens/trainable": 6820} +{"epoch": 0.0625, "grad_norm": 1.352664828300476, "learning_rate": 6e-05, "loss": 0.0500621572136879, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.05134, "step": 16, "tokens/total": 483504, "tokens/train_per_sec_per_gpu": 31.25, "tokens/trainable": 7258} +{"epoch": 0.06640625, "grad_norm": 1.773990273475647, "learning_rate": 6.400000000000001e-05, "loss": 0.060637038201093674, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.06251, "step": 17, "tokens/total": 514032, "tokens/train_per_sec_per_gpu": 35.72, "tokens/trainable": 7710} +{"epoch": 0.0703125, "grad_norm": 0.7572081089019775, "learning_rate": 6.800000000000001e-05, "loss": 0.028204206377267838, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.02861, "step": 18, "tokens/total": 544432, "tokens/train_per_sec_per_gpu": 32.19, "tokens/trainable": 8174} +{"epoch": 0.07421875, "grad_norm": 1.1870062351226807, "learning_rate": 7.2e-05, "loss": 0.03518473729491234, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03581, "step": 19, "tokens/total": 574880, "tokens/train_per_sec_per_gpu": 34.95, "tokens/trainable": 8617} +{"epoch": 0.078125, "grad_norm": 1.1347684860229492, "learning_rate": 7.6e-05, "loss": 0.03700350970029831, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0377, "step": 20, "tokens/total": 605408, "tokens/train_per_sec_per_gpu": 30.79, "tokens/trainable": 9037} +{"epoch": 0.08203125, "grad_norm": 0.6059332489967346, "learning_rate": 8e-05, "loss": 0.027005337178707123, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02737, "step": 21, "tokens/total": 635584, "tokens/train_per_sec_per_gpu": 37.21, "tokens/trainable": 9503} +{"epoch": 0.0859375, "grad_norm": 0.7608087658882141, "learning_rate": 8.4e-05, "loss": 0.025751473382115364, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.02609, "step": 22, "tokens/total": 665952, "tokens/train_per_sec_per_gpu": 37.31, "tokens/trainable": 9972} +{"epoch": 0.08984375, "grad_norm": 0.6488645672798157, "learning_rate": 8.800000000000001e-05, "loss": 0.021069440990686417, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02129, "step": 23, "tokens/total": 696240, "tokens/train_per_sec_per_gpu": 38.09, "tokens/trainable": 10447} +{"epoch": 0.09375, "grad_norm": 1.771577000617981, "learning_rate": 9.200000000000001e-05, "loss": 0.023897401988506317, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.02419, "step": 24, "tokens/total": 726464, "tokens/train_per_sec_per_gpu": 37.74, "tokens/trainable": 10899} +{"epoch": 0.09765625, "grad_norm": 1.4801814556121826, "learning_rate": 9.6e-05, "loss": 0.0472269207239151, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.04836, "step": 25, "tokens/total": 756704, "tokens/train_per_sec_per_gpu": 33.55, "tokens/trainable": 11360} +{"epoch": 0.1015625, "grad_norm": 1.178686261177063, "learning_rate": 0.0001, "loss": 0.031937334686517715, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.03245, "step": 26, "tokens/total": 786912, "tokens/train_per_sec_per_gpu": 29.74, "tokens/trainable": 11775} +{"epoch": 0.10546875, "grad_norm": 0.47292935848236084, "learning_rate": 9.99990636831587e-05, "loss": 0.008395236916840076, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.46, "memory/max_allocated (GiB)": 33.46, "ppl": 1.00843, "step": 27, "tokens/total": 815424, "tokens/train_per_sec_per_gpu": 40.47, "tokens/trainable": 12242} +{"epoch": 0.109375, "grad_norm": 0.7506940364837646, "learning_rate": 9.999625477159879e-05, "loss": 0.014507504180073738, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01461, "step": 28, "tokens/total": 845584, "tokens/train_per_sec_per_gpu": 34.07, "tokens/trainable": 12696} +{"epoch": 0.11328125, "grad_norm": 0.8390527963638306, "learning_rate": 9.999157338221051e-05, "loss": 0.020566720515489578, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.02078, "step": 29, "tokens/total": 875808, "tokens/train_per_sec_per_gpu": 34.89, "tokens/trainable": 13153} +{"epoch": 0.1171875, "grad_norm": 1.1266279220581055, "learning_rate": 9.998501970980562e-05, "loss": 0.010372643359005451, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01043, "step": 30, "tokens/total": 904096, "tokens/train_per_sec_per_gpu": 40.55, "tokens/trainable": 13624} +{"epoch": 0.12109375, "grad_norm": 0.6955512166023254, "learning_rate": 9.997659402710915e-05, "loss": 0.028533853590488434, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.02894, "step": 31, "tokens/total": 934400, "tokens/train_per_sec_per_gpu": 34.54, "tokens/trainable": 14064} +{"epoch": 0.125, "grad_norm": 0.5479819178581238, "learning_rate": 9.996629668474818e-05, "loss": 0.033125296235084534, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.03368, "step": 32, "tokens/total": 964832, "tokens/train_per_sec_per_gpu": 39.48, "tokens/trainable": 14565} +{"epoch": 0.12890625, "grad_norm": 0.5344933271408081, "learning_rate": 9.995412811123711e-05, "loss": 0.029415277764201164, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.02985, "step": 33, "tokens/total": 995040, "tokens/train_per_sec_per_gpu": 34.18, "tokens/trainable": 15005} +{"epoch": 0.1328125, "grad_norm": 0.6400431990623474, "learning_rate": 9.994008881295999e-05, "loss": 0.03012850694358349, "memory/device_reserved (GiB)": 34.72, "memory/max_active (GiB)": 33.64, "memory/max_allocated (GiB)": 33.64, "ppl": 1.03059, "step": 34, "tokens/total": 1024896, "tokens/train_per_sec_per_gpu": 32.51, "tokens/trainable": 15459} +{"epoch": 0.13671875, "grad_norm": 0.35772913694381714, "learning_rate": 9.992417937414932e-05, "loss": 0.02575090527534485, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02609, "step": 35, "tokens/total": 1054992, "tokens/train_per_sec_per_gpu": 38.65, "tokens/trainable": 15960} +{"epoch": 0.140625, "grad_norm": 0.7534306645393372, "learning_rate": 9.99064004568618e-05, "loss": 0.02033824287354946, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.02055, "step": 36, "tokens/total": 1085280, "tokens/train_per_sec_per_gpu": 34.98, "tokens/trainable": 16428} +{"epoch": 0.14453125, "grad_norm": 0.5049324035644531, "learning_rate": 9.988675280095074e-05, "loss": 0.019846521317958832, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.02004, "step": 37, "tokens/total": 1115504, "tokens/train_per_sec_per_gpu": 32.25, "tokens/trainable": 16884} +{"epoch": 0.1484375, "grad_norm": 0.4752073585987091, "learning_rate": 9.986523722403528e-05, "loss": 0.02174941822886467, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02199, "step": 38, "tokens/total": 1145712, "tokens/train_per_sec_per_gpu": 32.45, "tokens/trainable": 17341} +{"epoch": 0.15234375, "grad_norm": 0.7947481870651245, "learning_rate": 9.984185462146642e-05, "loss": 0.029066570103168488, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.02949, "step": 39, "tokens/total": 1176128, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 17786} +{"epoch": 0.15625, "grad_norm": 0.3997499942779541, "learning_rate": 9.98166059662897e-05, "loss": 0.017547229304909706, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0177, "step": 40, "tokens/total": 1206576, "tokens/train_per_sec_per_gpu": 35.31, "tokens/trainable": 18262} +{"epoch": 0.16015625, "grad_norm": 0.43478822708129883, "learning_rate": 9.978949230920472e-05, "loss": 0.008457450196146965, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00849, "step": 41, "tokens/total": 1236848, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 18716} +{"epoch": 0.1640625, "grad_norm": 0.5029982328414917, "learning_rate": 9.976051477852141e-05, "loss": 0.014029297046363354, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01413, "step": 42, "tokens/total": 1267088, "tokens/train_per_sec_per_gpu": 33.28, "tokens/trainable": 19157} +{"epoch": 0.16796875, "grad_norm": 1.298737645149231, "learning_rate": 9.972967458011312e-05, "loss": 0.03651078790426254, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.03719, "step": 43, "tokens/total": 1297360, "tokens/train_per_sec_per_gpu": 36.85, "tokens/trainable": 19647} +{"epoch": 0.171875, "grad_norm": 0.5446491241455078, "learning_rate": 9.96969729973664e-05, "loss": 0.01632726565003395, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01646, "step": 44, "tokens/total": 1327744, "tokens/train_per_sec_per_gpu": 35.37, "tokens/trainable": 20119} +{"epoch": 0.17578125, "grad_norm": 1.052590012550354, "learning_rate": 9.966241139112754e-05, "loss": 0.007975773885846138, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00801, "step": 45, "tokens/total": 1358096, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 20560} +{"epoch": 0.1796875, "grad_norm": 0.2636128067970276, "learning_rate": 9.96259911996461e-05, "loss": 0.004431102424860001, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00444, "step": 46, "tokens/total": 1388240, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 20992} +{"epoch": 0.18359375, "grad_norm": 0.8592759966850281, "learning_rate": 9.958771393851491e-05, "loss": 0.02455780655145645, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.26, "memory/max_allocated (GiB)": 33.26, "ppl": 1.02486, "step": 47, "tokens/total": 1416240, "tokens/train_per_sec_per_gpu": 31.97, "tokens/trainable": 21441} +{"epoch": 0.1875, "grad_norm": 1.0893580913543701, "learning_rate": 9.954758120060702e-05, "loss": 0.016398968175053596, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.01653, "step": 48, "tokens/total": 1446880, "tokens/train_per_sec_per_gpu": 34.17, "tokens/trainable": 21901} +{"epoch": 0.19140625, "grad_norm": 0.6647804379463196, "learning_rate": 9.950559465600948e-05, "loss": 0.007302718702703714, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00733, "step": 49, "tokens/total": 1477168, "tokens/train_per_sec_per_gpu": 33.72, "tokens/trainable": 22341} +{"epoch": 0.1953125, "grad_norm": 0.2949844300746918, "learning_rate": 9.946175605195379e-05, "loss": 0.007496782578527927, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00752, "step": 50, "tokens/total": 1507712, "tokens/train_per_sec_per_gpu": 33.56, "tokens/trainable": 22833} +{"epoch": 0.19921875, "grad_norm": 0.709971010684967, "learning_rate": 9.941606721274322e-05, "loss": 0.008996322751045227, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00904, "step": 51, "tokens/total": 1537936, "tokens/train_per_sec_per_gpu": 31.07, "tokens/trainable": 23277} +{"epoch": 0.203125, "grad_norm": 2.150784969329834, "learning_rate": 9.936853003967685e-05, "loss": 0.02703060768544674, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0274, "step": 52, "tokens/total": 1568352, "tokens/train_per_sec_per_gpu": 36.11, "tokens/trainable": 23759} +{"epoch": 0.20703125, "grad_norm": 0.9372985363006592, "learning_rate": 9.93191465109705e-05, "loss": 0.043413519859313965, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.04437, "step": 53, "tokens/total": 1598992, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 24193} +{"epoch": 0.2109375, "grad_norm": 1.728373408317566, "learning_rate": 9.926791868167438e-05, "loss": 0.027337459847331047, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.02771, "step": 54, "tokens/total": 1629488, "tokens/train_per_sec_per_gpu": 34.01, "tokens/trainable": 24619} +{"epoch": 0.21484375, "grad_norm": 0.9787118434906006, "learning_rate": 9.921484868358753e-05, "loss": 0.024128694087266922, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.02442, "step": 55, "tokens/total": 1659696, "tokens/train_per_sec_per_gpu": 33.23, "tokens/trainable": 25075} +{"epoch": 0.21875, "grad_norm": 0.5342002511024475, "learning_rate": 9.915993872516924e-05, "loss": 0.0141680296510458, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.01427, "step": 56, "tokens/total": 1689872, "tokens/train_per_sec_per_gpu": 31.96, "tokens/trainable": 25519} +{"epoch": 0.22265625, "grad_norm": 0.34215787053108215, "learning_rate": 9.9103191091447e-05, "loss": 0.007503957487642765, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00753, "step": 57, "tokens/total": 1720144, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 25966} +{"epoch": 0.2265625, "grad_norm": 0.3897297978401184, "learning_rate": 9.904460814392147e-05, "loss": 0.0075007774867117405, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00753, "step": 58, "tokens/total": 1750448, "tokens/train_per_sec_per_gpu": 34.38, "tokens/trainable": 26423} +{"epoch": 0.23046875, "grad_norm": 2.039067268371582, "learning_rate": 9.898419232046825e-05, "loss": 0.023454945534467697, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.02373, "step": 59, "tokens/total": 1781088, "tokens/train_per_sec_per_gpu": 38.99, "tokens/trainable": 26934} +{"epoch": 0.234375, "grad_norm": 0.5912706851959229, "learning_rate": 9.892194613523633e-05, "loss": 0.006885843351483345, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00691, "step": 60, "tokens/total": 1811344, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 27393} +{"epoch": 0.23828125, "grad_norm": 0.6040364503860474, "learning_rate": 9.885787217854357e-05, "loss": 0.004220158793032169, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00423, "step": 61, "tokens/total": 1841600, "tokens/train_per_sec_per_gpu": 33.4, "tokens/trainable": 27852} +{"epoch": 0.2421875, "grad_norm": 0.8806986808776855, "learning_rate": 9.879197311676887e-05, "loss": 0.018320046365261078, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01849, "step": 62, "tokens/total": 1872048, "tokens/train_per_sec_per_gpu": 33.36, "tokens/trainable": 28292} +{"epoch": 0.24609375, "grad_norm": 1.204670786857605, "learning_rate": 9.872425169224113e-05, "loss": 0.028115103021264076, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.02851, "step": 63, "tokens/total": 1902592, "tokens/train_per_sec_per_gpu": 37.85, "tokens/trainable": 28798} +{"epoch": 0.25, "grad_norm": 1.5364751815795898, "learning_rate": 9.865471072312528e-05, "loss": 0.051395732909440994, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.05274, "step": 64, "tokens/total": 1933088, "tokens/train_per_sec_per_gpu": 36.24, "tokens/trainable": 29283} +{"epoch": 0.25390625, "grad_norm": 2.3558332920074463, "learning_rate": 9.858335310330492e-05, "loss": 0.01967993751168251, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.01987, "step": 65, "tokens/total": 1963296, "tokens/train_per_sec_per_gpu": 32.35, "tokens/trainable": 29745} +{"epoch": 0.2578125, "grad_norm": 1.269558072090149, "learning_rate": 9.851018180226185e-05, "loss": 0.017027614638209343, "memory/device_reserved (GiB)": 34.87, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01717, "step": 66, "tokens/total": 1993760, "tokens/train_per_sec_per_gpu": 31.64, "tokens/trainable": 30186} +{"epoch": 0.26171875, "grad_norm": 0.287238746881485, "learning_rate": 9.843519986495259e-05, "loss": 0.008575644344091415, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00861, "step": 67, "tokens/total": 2024144, "tokens/train_per_sec_per_gpu": 33.7, "tokens/trainable": 30622} +{"epoch": 0.265625, "grad_norm": 0.30087289214134216, "learning_rate": 9.835841041168162e-05, "loss": 0.013771848753094673, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.01387, "step": 68, "tokens/total": 2054608, "tokens/train_per_sec_per_gpu": 35.93, "tokens/trainable": 31097} +{"epoch": 0.26953125, "grad_norm": 0.2974952459335327, "learning_rate": 9.82798166379715e-05, "loss": 0.008139240555465221, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00817, "step": 69, "tokens/total": 2084912, "tokens/train_per_sec_per_gpu": 37.0, "tokens/trainable": 31595} +{"epoch": 0.2734375, "grad_norm": 1.095276117324829, "learning_rate": 9.819942181443002e-05, "loss": 0.020748011767864227, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02096, "step": 70, "tokens/total": 2115216, "tokens/train_per_sec_per_gpu": 31.08, "tokens/trainable": 32036} +{"epoch": 0.27734375, "grad_norm": 0.4999067485332489, "learning_rate": 9.811722928661392e-05, "loss": 0.0232261773198843, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0235, "step": 71, "tokens/total": 2145424, "tokens/train_per_sec_per_gpu": 28.43, "tokens/trainable": 32428} +{"epoch": 0.28125, "grad_norm": 0.2014380246400833, "learning_rate": 9.803324247488975e-05, "loss": 0.004090111702680588, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0041, "step": 72, "tokens/total": 2175648, "tokens/train_per_sec_per_gpu": 32.55, "tokens/trainable": 32880} +{"epoch": 0.28515625, "grad_norm": 0.24525801837444305, "learning_rate": 9.794746487429161e-05, "loss": 0.004098955541849136, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00411, "step": 73, "tokens/total": 2205856, "tokens/train_per_sec_per_gpu": 33.95, "tokens/trainable": 33323} +{"epoch": 0.2890625, "grad_norm": 0.26871851086616516, "learning_rate": 9.785990005437554e-05, "loss": 0.008776037022471428, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00881, "step": 74, "tokens/total": 2236352, "tokens/train_per_sec_per_gpu": 36.31, "tokens/trainable": 33792} +{"epoch": 0.29296875, "grad_norm": 0.4705350697040558, "learning_rate": 9.777055165907117e-05, "loss": 0.025479601696133614, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.02581, "step": 75, "tokens/total": 2266480, "tokens/train_per_sec_per_gpu": 35.72, "tokens/trainable": 34223} +{"epoch": 0.296875, "grad_norm": 0.7641831636428833, "learning_rate": 9.767942340652993e-05, "loss": 0.03422342985868454, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.03482, "step": 76, "tokens/total": 2296496, "tokens/train_per_sec_per_gpu": 29.96, "tokens/trainable": 34649} +{"epoch": 0.30078125, "grad_norm": 0.490901380777359, "learning_rate": 9.758651908897035e-05, "loss": 0.019773798063397408, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01997, "step": 77, "tokens/total": 2327040, "tokens/train_per_sec_per_gpu": 36.08, "tokens/trainable": 35094} +{"epoch": 0.3046875, "grad_norm": 0.6036896109580994, "learning_rate": 9.749184257252033e-05, "loss": 0.029439125210046768, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.02988, "step": 78, "tokens/total": 2357328, "tokens/train_per_sec_per_gpu": 32.2, "tokens/trainable": 35534} +{"epoch": 0.30859375, "grad_norm": 0.5037710666656494, "learning_rate": 9.739539779705614e-05, "loss": 0.013313151895999908, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0134, "step": 79, "tokens/total": 2387664, "tokens/train_per_sec_per_gpu": 36.22, "tokens/trainable": 36029} +{"epoch": 0.3125, "grad_norm": 0.4796348214149475, "learning_rate": 9.729718877603861e-05, "loss": 0.02049200050532818, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0207, "step": 80, "tokens/total": 2418000, "tokens/train_per_sec_per_gpu": 36.0, "tokens/trainable": 36469} +{"epoch": 0.31640625, "grad_norm": 0.9642574191093445, "learning_rate": 9.719721959634592e-05, "loss": 0.04148370027542114, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.04236, "step": 81, "tokens/total": 2448720, "tokens/train_per_sec_per_gpu": 31.16, "tokens/trainable": 36906} +{"epoch": 0.3203125, "grad_norm": 0.22247610986232758, "learning_rate": 9.709549441810375e-05, "loss": 0.007624159567058086, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00765, "step": 82, "tokens/total": 2479248, "tokens/train_per_sec_per_gpu": 39.43, "tokens/trainable": 37390} +{"epoch": 0.32421875, "grad_norm": 0.6237291693687439, "learning_rate": 9.699201747451195e-05, "loss": 0.027217810973525047, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.02759, "step": 83, "tokens/total": 2509408, "tokens/train_per_sec_per_gpu": 31.17, "tokens/trainable": 37838} +{"epoch": 0.328125, "grad_norm": 0.2244887351989746, "learning_rate": 9.688679307166854e-05, "loss": 0.009322088211774826, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00937, "step": 84, "tokens/total": 2539776, "tokens/train_per_sec_per_gpu": 35.56, "tokens/trainable": 38310} +{"epoch": 0.33203125, "grad_norm": 0.4746335446834564, "learning_rate": 9.677982558839042e-05, "loss": 0.027634991332888603, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.02802, "step": 85, "tokens/total": 2569840, "tokens/train_per_sec_per_gpu": 34.52, "tokens/trainable": 38789} +{"epoch": 0.3359375, "grad_norm": 0.33432865142822266, "learning_rate": 9.66711194760312e-05, "loss": 0.017999891191720963, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01816, "step": 86, "tokens/total": 2600192, "tokens/train_per_sec_per_gpu": 36.83, "tokens/trainable": 39277} +{"epoch": 0.33984375, "grad_norm": 0.5810937285423279, "learning_rate": 9.656067925829593e-05, "loss": 0.02710646390914917, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.02748, "step": 87, "tokens/total": 2630640, "tokens/train_per_sec_per_gpu": 36.17, "tokens/trainable": 39737} +{"epoch": 0.34375, "grad_norm": 0.23032420873641968, "learning_rate": 9.644850953105288e-05, "loss": 0.007609693333506584, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00764, "step": 88, "tokens/total": 2660832, "tokens/train_per_sec_per_gpu": 29.98, "tokens/trainable": 40179} +{"epoch": 0.34765625, "grad_norm": 0.2694112956523895, "learning_rate": 9.633461496214225e-05, "loss": 0.010818381793797016, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01088, "step": 89, "tokens/total": 2691328, "tokens/train_per_sec_per_gpu": 33.05, "tokens/trainable": 40649} +{"epoch": 0.3515625, "grad_norm": 0.8180995583534241, "learning_rate": 9.621900029118195e-05, "loss": 0.02754024788737297, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.41, "memory/max_allocated (GiB)": 33.41, "ppl": 1.02792, "step": 90, "tokens/total": 2719696, "tokens/train_per_sec_per_gpu": 31.94, "tokens/trainable": 41080} +{"epoch": 0.35546875, "grad_norm": 0.1282704770565033, "learning_rate": 9.610167032937036e-05, "loss": 0.0028113755397498608, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00282, "step": 91, "tokens/total": 2750272, "tokens/train_per_sec_per_gpu": 38.54, "tokens/trainable": 41599} +{"epoch": 0.359375, "grad_norm": 0.28648310899734497, "learning_rate": 9.598262995928611e-05, "loss": 0.010297400876879692, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.01035, "step": 92, "tokens/total": 2780656, "tokens/train_per_sec_per_gpu": 37.25, "tokens/trainable": 42076} +{"epoch": 0.36328125, "grad_norm": 0.18042024970054626, "learning_rate": 9.586188413468492e-05, "loss": 0.0049249883741140366, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00494, "step": 93, "tokens/total": 2811088, "tokens/train_per_sec_per_gpu": 34.12, "tokens/trainable": 42522} +{"epoch": 0.3671875, "grad_norm": 2.636540174484253, "learning_rate": 9.57394378802934e-05, "loss": 0.014661800116300583, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01477, "step": 94, "tokens/total": 2841520, "tokens/train_per_sec_per_gpu": 34.81, "tokens/trainable": 42974} +{"epoch": 0.37109375, "grad_norm": 1.0461680889129639, "learning_rate": 9.56152962916e-05, "loss": 0.01939905434846878, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.01959, "step": 95, "tokens/total": 2871600, "tokens/train_per_sec_per_gpu": 31.12, "tokens/trainable": 43380} +{"epoch": 0.375, "grad_norm": 0.1774737685918808, "learning_rate": 9.548946453464296e-05, "loss": 0.005020626820623875, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00503, "step": 96, "tokens/total": 2902144, "tokens/train_per_sec_per_gpu": 34.54, "tokens/trainable": 43865} +{"epoch": 0.37890625, "grad_norm": 0.26704758405685425, "learning_rate": 9.53619478457953e-05, "loss": 0.014278757385909557, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01438, "step": 97, "tokens/total": 2932272, "tokens/train_per_sec_per_gpu": 32.29, "tokens/trainable": 44328} +{"epoch": 0.3828125, "grad_norm": 0.15346744656562805, "learning_rate": 9.523275153154695e-05, "loss": 0.0042286161333322525, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00424, "step": 98, "tokens/total": 2962032, "tokens/train_per_sec_per_gpu": 31.36, "tokens/trainable": 44778} +{"epoch": 0.38671875, "grad_norm": 0.3834576904773712, "learning_rate": 9.51018809682839e-05, "loss": 0.011703035794198513, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01177, "step": 99, "tokens/total": 2992256, "tokens/train_per_sec_per_gpu": 37.68, "tokens/trainable": 45225} +{"epoch": 0.390625, "grad_norm": 0.7086752653121948, "learning_rate": 9.49693416020645e-05, "loss": 0.014463641680777073, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01457, "step": 100, "tokens/total": 3022608, "tokens/train_per_sec_per_gpu": 36.26, "tokens/trainable": 45678} +{"epoch": 0.39453125, "grad_norm": 0.27034133672714233, "learning_rate": 9.483513894839276e-05, "loss": 0.00494391517713666, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00496, "step": 101, "tokens/total": 3052992, "tokens/train_per_sec_per_gpu": 33.19, "tokens/trainable": 46125} +{"epoch": 0.3984375, "grad_norm": 0.17598937451839447, "learning_rate": 9.469927859198888e-05, "loss": 0.002658358309417963, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00266, "step": 102, "tokens/total": 3083392, "tokens/train_per_sec_per_gpu": 40.22, "tokens/trainable": 46612} +{"epoch": 0.40234375, "grad_norm": 0.25017818808555603, "learning_rate": 9.456176618655689e-05, "loss": 0.005640048533678055, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00566, "step": 103, "tokens/total": 3113744, "tokens/train_per_sec_per_gpu": 33.0, "tokens/trainable": 47059} +{"epoch": 0.40625, "grad_norm": 0.7971018552780151, "learning_rate": 9.442260745454927e-05, "loss": 0.03405749425292015, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.03464, "step": 104, "tokens/total": 3144272, "tokens/train_per_sec_per_gpu": 34.48, "tokens/trainable": 47536} +{"epoch": 0.41015625, "grad_norm": 0.3616940677165985, "learning_rate": 9.428180818692884e-05, "loss": 0.004266355652362108, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00428, "step": 105, "tokens/total": 3174512, "tokens/train_per_sec_per_gpu": 34.44, "tokens/trainable": 48037} +{"epoch": 0.4140625, "grad_norm": 0.37930840253829956, "learning_rate": 9.413937424292791e-05, "loss": 0.018286891281604767, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01846, "step": 106, "tokens/total": 3204896, "tokens/train_per_sec_per_gpu": 32.99, "tokens/trainable": 48491} +{"epoch": 0.41796875, "grad_norm": 0.2938750088214874, "learning_rate": 9.399531154980424e-05, "loss": 0.007349148392677307, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.00738, "step": 107, "tokens/total": 3235360, "tokens/train_per_sec_per_gpu": 33.38, "tokens/trainable": 48940} +{"epoch": 0.421875, "grad_norm": 0.7292595505714417, "learning_rate": 9.384962610259455e-05, "loss": 0.008774684742093086, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00881, "step": 108, "tokens/total": 3265552, "tokens/train_per_sec_per_gpu": 32.07, "tokens/trainable": 49389} +{"epoch": 0.42578125, "grad_norm": 0.14338180422782898, "learning_rate": 9.370232396386494e-05, "loss": 0.003269340842962265, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.00327, "step": 109, "tokens/total": 3293856, "tokens/train_per_sec_per_gpu": 36.45, "tokens/trainable": 49838} +{"epoch": 0.4296875, "grad_norm": 0.5673946142196655, "learning_rate": 9.355341126345868e-05, "loss": 0.015835193917155266, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.01596, "step": 110, "tokens/total": 3323984, "tokens/train_per_sec_per_gpu": 34.34, "tokens/trainable": 50310} +{"epoch": 0.43359375, "grad_norm": 0.4097978174686432, "learning_rate": 9.340289419824107e-05, "loss": 0.015539250336587429, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01566, "step": 111, "tokens/total": 3354320, "tokens/train_per_sec_per_gpu": 34.04, "tokens/trainable": 50755} +{"epoch": 0.4375, "grad_norm": 0.2639133036136627, "learning_rate": 9.325077903184159e-05, "loss": 0.010360169224441051, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01041, "step": 112, "tokens/total": 3384880, "tokens/train_per_sec_per_gpu": 31.99, "tokens/trainable": 51213} +{"epoch": 0.44140625, "grad_norm": 0.03919512405991554, "learning_rate": 9.30970720943932e-05, "loss": 0.000984038575552404, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00098, "step": 113, "tokens/total": 3415104, "tokens/train_per_sec_per_gpu": 34.18, "tokens/trainable": 51660} +{"epoch": 0.4453125, "grad_norm": 0.517244815826416, "learning_rate": 9.2941779782269e-05, "loss": 0.014096877537667751, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0142, "step": 114, "tokens/total": 3445504, "tokens/train_per_sec_per_gpu": 32.73, "tokens/trainable": 52133} +{"epoch": 0.44921875, "grad_norm": 0.38755857944488525, "learning_rate": 9.278490855781596e-05, "loss": 0.010432498529553413, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01049, "step": 115, "tokens/total": 3475744, "tokens/train_per_sec_per_gpu": 37.9, "tokens/trainable": 52580} +{"epoch": 0.453125, "grad_norm": 0.5933223366737366, "learning_rate": 9.262646494908604e-05, "loss": 0.018632538616657257, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01881, "step": 116, "tokens/total": 3506016, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 53033} +{"epoch": 0.45703125, "grad_norm": 1.0803433656692505, "learning_rate": 9.246645554956457e-05, "loss": 0.018085651099681854, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01825, "step": 117, "tokens/total": 3536256, "tokens/train_per_sec_per_gpu": 35.78, "tokens/trainable": 53517} +{"epoch": 0.4609375, "grad_norm": 0.18590858578681946, "learning_rate": 9.230488701789578e-05, "loss": 0.0024977584835141897, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.0025, "step": 118, "tokens/total": 3566480, "tokens/train_per_sec_per_gpu": 35.08, "tokens/trainable": 53967} +{"epoch": 0.46484375, "grad_norm": 0.20355315506458282, "learning_rate": 9.214176607760577e-05, "loss": 0.0031010895036160946, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00311, "step": 119, "tokens/total": 3596624, "tokens/train_per_sec_per_gpu": 35.81, "tokens/trainable": 54430} +{"epoch": 0.46875, "grad_norm": 0.12187723815441132, "learning_rate": 9.197709951682268e-05, "loss": 0.0025250515900552273, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00253, "step": 120, "tokens/total": 3627168, "tokens/train_per_sec_per_gpu": 35.77, "tokens/trainable": 54896} +{"epoch": 0.47265625, "grad_norm": 0.4177272915840149, "learning_rate": 9.181089418799428e-05, "loss": 0.01381619367748499, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.01391, "step": 121, "tokens/total": 3657632, "tokens/train_per_sec_per_gpu": 38.27, "tokens/trainable": 55379} +{"epoch": 0.4765625, "grad_norm": 0.1338144689798355, "learning_rate": 9.164315700760271e-05, "loss": 0.002670921152457595, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00267, "step": 122, "tokens/total": 3687824, "tokens/train_per_sec_per_gpu": 32.2, "tokens/trainable": 55803} +{"epoch": 0.48046875, "grad_norm": 0.2588568329811096, "learning_rate": 9.147389495587671e-05, "loss": 0.005417007952928543, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00543, "step": 123, "tokens/total": 3718320, "tokens/train_per_sec_per_gpu": 36.99, "tokens/trainable": 56249} +{"epoch": 0.484375, "grad_norm": 0.21248272061347961, "learning_rate": 9.130311507650116e-05, "loss": 0.002708941698074341, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00271, "step": 124, "tokens/total": 3748672, "tokens/train_per_sec_per_gpu": 32.85, "tokens/trainable": 56713} +{"epoch": 0.48828125, "grad_norm": 0.344331294298172, "learning_rate": 9.113082447632394e-05, "loss": 0.00817079283297062, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0082, "step": 125, "tokens/total": 3778976, "tokens/train_per_sec_per_gpu": 39.62, "tokens/trainable": 57196} +{"epoch": 0.4921875, "grad_norm": 0.26439207792282104, "learning_rate": 9.09570303250602e-05, "loss": 0.005593656096607447, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00561, "step": 126, "tokens/total": 3809296, "tokens/train_per_sec_per_gpu": 37.02, "tokens/trainable": 57680} +{"epoch": 0.49609375, "grad_norm": 0.11266636103391647, "learning_rate": 9.078173985499394e-05, "loss": 0.0022880970500409603, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00229, "step": 127, "tokens/total": 3839328, "tokens/train_per_sec_per_gpu": 32.3, "tokens/trainable": 58110} +{"epoch": 0.5, "grad_norm": 0.11059004068374634, "learning_rate": 9.060496036067713e-05, "loss": 0.0021013133227825165, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0021, "step": 128, "tokens/total": 3869824, "tokens/train_per_sec_per_gpu": 32.16, "tokens/trainable": 58534} +{"epoch": 0.50390625, "grad_norm": 0.2581784129142761, "learning_rate": 9.042669919862615e-05, "loss": 0.010128125548362732, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01018, "step": 129, "tokens/total": 3900080, "tokens/train_per_sec_per_gpu": 35.6, "tokens/trainable": 58998} +{"epoch": 0.5078125, "grad_norm": 0.2631431818008423, "learning_rate": 9.024696378701557e-05, "loss": 0.009796728380024433, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00984, "step": 130, "tokens/total": 3930560, "tokens/train_per_sec_per_gpu": 32.2, "tokens/trainable": 59448} +{"epoch": 0.51171875, "grad_norm": 0.08005515486001968, "learning_rate": 9.006576160536948e-05, "loss": 0.0013428543461486697, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00134, "step": 131, "tokens/total": 3960496, "tokens/train_per_sec_per_gpu": 32.45, "tokens/trainable": 59906} +{"epoch": 0.515625, "grad_norm": 0.2250908762216568, "learning_rate": 8.988310019425035e-05, "loss": 0.0020651742815971375, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00207, "step": 132, "tokens/total": 3990928, "tokens/train_per_sec_per_gpu": 34.72, "tokens/trainable": 60350} +{"epoch": 0.51953125, "grad_norm": 0.25850772857666016, "learning_rate": 8.969898715494506e-05, "loss": 0.004448378924280405, "memory/device_reserved (GiB)": 37.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00446, "step": 133, "tokens/total": 4021264, "tokens/train_per_sec_per_gpu": 36.61, "tokens/trainable": 60831} +{"epoch": 0.5234375, "grad_norm": 0.4923039972782135, "learning_rate": 8.951343014914869e-05, "loss": 0.01576503925025463, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01589, "step": 134, "tokens/total": 4051728, "tokens/train_per_sec_per_gpu": 34.46, "tokens/trainable": 61305} +{"epoch": 0.52734375, "grad_norm": 0.019552618265151978, "learning_rate": 8.932643689864568e-05, "loss": 0.0003579487092792988, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00036, "step": 135, "tokens/total": 4081920, "tokens/train_per_sec_per_gpu": 38.09, "tokens/trainable": 61792} +{"epoch": 0.53125, "grad_norm": 0.08227464556694031, "learning_rate": 8.913801518498845e-05, "loss": 0.001130630262196064, "memory/device_reserved (GiB)": 38.55, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00113, "step": 136, "tokens/total": 4112304, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 62253} +{"epoch": 0.53515625, "grad_norm": 0.4993005394935608, "learning_rate": 8.894817284917364e-05, "loss": 0.0084285419434309, "memory/device_reserved (GiB)": 38.55, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00846, "step": 137, "tokens/total": 4142512, "tokens/train_per_sec_per_gpu": 32.04, "tokens/trainable": 62707} +{"epoch": 0.5390625, "grad_norm": 0.10504501312971115, "learning_rate": 8.875691779131569e-05, "loss": 0.0017363994847983122, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00174, "step": 138, "tokens/total": 4172992, "tokens/train_per_sec_per_gpu": 29.69, "tokens/trainable": 63145} +{"epoch": 0.54296875, "grad_norm": 0.34308046102523804, "learning_rate": 8.856425797031829e-05, "loss": 0.0045009395107626915, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00451, "step": 139, "tokens/total": 4203136, "tokens/train_per_sec_per_gpu": 34.39, "tokens/trainable": 63587} +{"epoch": 0.546875, "grad_norm": 0.4159857928752899, "learning_rate": 8.837020140354295e-05, "loss": 0.007854020223021507, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00788, "step": 140, "tokens/total": 4233456, "tokens/train_per_sec_per_gpu": 37.25, "tokens/trainable": 64052} +{"epoch": 0.55078125, "grad_norm": 0.6004462242126465, "learning_rate": 8.817475616647554e-05, "loss": 0.013760390691459179, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01386, "step": 141, "tokens/total": 4263728, "tokens/train_per_sec_per_gpu": 34.81, "tokens/trainable": 64526} +{"epoch": 0.5546875, "grad_norm": 0.011325858533382416, "learning_rate": 8.797793039239017e-05, "loss": 0.00024004181614145637, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.00024, "step": 142, "tokens/total": 4294432, "tokens/train_per_sec_per_gpu": 35.15, "tokens/trainable": 64983} +{"epoch": 0.55859375, "grad_norm": 0.5004990100860596, "learning_rate": 8.777973227201069e-05, "loss": 0.01317769568413496, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01326, "step": 143, "tokens/total": 4324960, "tokens/train_per_sec_per_gpu": 38.0, "tokens/trainable": 65492} +{"epoch": 0.5625, "grad_norm": 0.9903383255004883, "learning_rate": 8.758017005316988e-05, "loss": 0.0297666247934103, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.03021, "step": 144, "tokens/total": 4355424, "tokens/train_per_sec_per_gpu": 36.11, "tokens/trainable": 65925} +{"epoch": 0.56640625, "grad_norm": 0.21414855122566223, "learning_rate": 8.737925204046629e-05, "loss": 0.009566199965775013, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00961, "step": 145, "tokens/total": 4385712, "tokens/train_per_sec_per_gpu": 31.78, "tokens/trainable": 66383} +{"epoch": 0.5703125, "grad_norm": 0.43486288189888, "learning_rate": 8.717698659491851e-05, "loss": 0.006363986060023308, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00638, "step": 146, "tokens/total": 4416016, "tokens/train_per_sec_per_gpu": 31.79, "tokens/trainable": 66840} +{"epoch": 0.57421875, "grad_norm": 0.4194156527519226, "learning_rate": 8.697338213361735e-05, "loss": 0.01065310649573803, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01071, "step": 147, "tokens/total": 4446368, "tokens/train_per_sec_per_gpu": 32.14, "tokens/trainable": 67297} +{"epoch": 0.578125, "grad_norm": 0.1669190227985382, "learning_rate": 8.676844712937552e-05, "loss": 0.006577126681804657, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0066, "step": 148, "tokens/total": 4476880, "tokens/train_per_sec_per_gpu": 37.88, "tokens/trainable": 67769} +{"epoch": 0.58203125, "grad_norm": 0.1995057761669159, "learning_rate": 8.656219011037509e-05, "loss": 0.006504375487565994, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00653, "step": 149, "tokens/total": 4507232, "tokens/train_per_sec_per_gpu": 33.2, "tokens/trainable": 68257} +{"epoch": 0.5859375, "grad_norm": 0.6449038982391357, "learning_rate": 8.63546196598125e-05, "loss": 0.020274966955184937, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02048, "step": 150, "tokens/total": 4537376, "tokens/train_per_sec_per_gpu": 35.92, "tokens/trainable": 68706} +{"epoch": 0.58984375, "grad_norm": 0.4044327437877655, "learning_rate": 8.614574441554145e-05, "loss": 0.012972041964530945, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01306, "step": 151, "tokens/total": 4567584, "tokens/train_per_sec_per_gpu": 33.7, "tokens/trainable": 69131} +{"epoch": 0.59375, "grad_norm": 0.45228511095046997, "learning_rate": 8.593557306971349e-05, "loss": 0.02865610271692276, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.02907, "step": 152, "tokens/total": 4597792, "tokens/train_per_sec_per_gpu": 32.34, "tokens/trainable": 69586} +{"epoch": 0.59765625, "grad_norm": 0.1369956135749817, "learning_rate": 8.572411436841618e-05, "loss": 0.004469756968319416, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00448, "step": 153, "tokens/total": 4627936, "tokens/train_per_sec_per_gpu": 33.19, "tokens/trainable": 70061} +{"epoch": 0.6015625, "grad_norm": 0.132548987865448, "learning_rate": 8.551137711130922e-05, "loss": 0.006225723773241043, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00625, "step": 154, "tokens/total": 4658352, "tokens/train_per_sec_per_gpu": 28.28, "tokens/trainable": 70467} +{"epoch": 0.60546875, "grad_norm": 0.15946446359157562, "learning_rate": 8.529737015125824e-05, "loss": 0.005463356152176857, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00548, "step": 155, "tokens/total": 4688896, "tokens/train_per_sec_per_gpu": 33.49, "tokens/trainable": 70933} +{"epoch": 0.609375, "grad_norm": 0.3434015214443207, "learning_rate": 8.508210239396639e-05, "loss": 0.014093048870563507, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01419, "step": 156, "tokens/total": 4719168, "tokens/train_per_sec_per_gpu": 34.69, "tokens/trainable": 71382} +{"epoch": 0.61328125, "grad_norm": 0.06986619532108307, "learning_rate": 8.486558279760375e-05, "loss": 0.0025533996522426605, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00256, "step": 157, "tokens/total": 4749136, "tokens/train_per_sec_per_gpu": 31.13, "tokens/trainable": 71808} +{"epoch": 0.6171875, "grad_norm": 0.34366294741630554, "learning_rate": 8.464782037243449e-05, "loss": 0.013532277196645737, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.01362, "step": 158, "tokens/total": 4779472, "tokens/train_per_sec_per_gpu": 33.23, "tokens/trainable": 72249} +{"epoch": 0.62109375, "grad_norm": 0.5110010504722595, "learning_rate": 8.442882418044202e-05, "loss": 0.01647217571735382, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.01661, "step": 159, "tokens/total": 4809632, "tokens/train_per_sec_per_gpu": 35.57, "tokens/trainable": 72726} +{"epoch": 0.625, "grad_norm": 0.10558311641216278, "learning_rate": 8.420860333495179e-05, "loss": 0.004920906387269497, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00493, "step": 160, "tokens/total": 4840112, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 73148} +{"epoch": 0.62890625, "grad_norm": 0.16564413905143738, "learning_rate": 8.398716700025208e-05, "loss": 0.00695970980450511, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 34.04, "memory/max_allocated (GiB)": 34.04, "ppl": 1.00698, "step": 161, "tokens/total": 4870656, "tokens/train_per_sec_per_gpu": 31.72, "tokens/trainable": 73600} +{"epoch": 0.6328125, "grad_norm": 0.28341594338417053, "learning_rate": 8.376452439121266e-05, "loss": 0.010892837308347225, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.01095, "step": 162, "tokens/total": 4900608, "tokens/train_per_sec_per_gpu": 36.1, "tokens/trainable": 74068} +{"epoch": 0.63671875, "grad_norm": 0.05444963648915291, "learning_rate": 8.354068477290124e-05, "loss": 0.0015460889553651214, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00155, "step": 163, "tokens/total": 4930752, "tokens/train_per_sec_per_gpu": 36.15, "tokens/trainable": 74527} +{"epoch": 0.640625, "grad_norm": 0.09271538257598877, "learning_rate": 8.331565746019807e-05, "loss": 0.0028422921895980835, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00285, "step": 164, "tokens/total": 4961008, "tokens/train_per_sec_per_gpu": 32.25, "tokens/trainable": 74992} +{"epoch": 0.64453125, "grad_norm": 0.20550787448883057, "learning_rate": 8.308945181740812e-05, "loss": 0.004469836596399546, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00448, "step": 165, "tokens/total": 4991200, "tokens/train_per_sec_per_gpu": 35.7, "tokens/trainable": 75442} +{"epoch": 0.6484375, "grad_norm": 0.3527662754058838, "learning_rate": 8.286207725787153e-05, "loss": 0.010943438857793808, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.011, "step": 166, "tokens/total": 5021600, "tokens/train_per_sec_per_gpu": 31.6, "tokens/trainable": 75887} +{"epoch": 0.65234375, "grad_norm": 0.4151509702205658, "learning_rate": 8.263354324357182e-05, "loss": 0.010490473359823227, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.01055, "step": 167, "tokens/total": 5052032, "tokens/train_per_sec_per_gpu": 34.48, "tokens/trainable": 76331} +{"epoch": 0.65625, "grad_norm": 0.13094273209571838, "learning_rate": 8.240385928474219e-05, "loss": 0.0015372417401522398, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00154, "step": 168, "tokens/total": 5080256, "tokens/train_per_sec_per_gpu": 36.24, "tokens/trainable": 76745} +{"epoch": 0.66015625, "grad_norm": 0.05917811021208763, "learning_rate": 8.217303493946967e-05, "loss": 0.0017156600952148438, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00172, "step": 169, "tokens/total": 5110784, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 77201} +{"epoch": 0.6640625, "grad_norm": 0.08870573341846466, "learning_rate": 8.194107981329746e-05, "loss": 0.0014357661129906774, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00144, "step": 170, "tokens/total": 5141200, "tokens/train_per_sec_per_gpu": 33.76, "tokens/trainable": 77655} +{"epoch": 0.66796875, "grad_norm": 0.16937553882598877, "learning_rate": 8.170800355882518e-05, "loss": 0.0023305683862417936, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00233, "step": 171, "tokens/total": 5171760, "tokens/train_per_sec_per_gpu": 39.68, "tokens/trainable": 78122} +{"epoch": 0.671875, "grad_norm": 0.1098952665925026, "learning_rate": 8.147381587530713e-05, "loss": 0.001408248208463192, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00141, "step": 172, "tokens/total": 5202272, "tokens/train_per_sec_per_gpu": 33.98, "tokens/trainable": 78576} +{"epoch": 0.67578125, "grad_norm": 0.20372426509857178, "learning_rate": 8.123852650824877e-05, "loss": 0.005307321436703205, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00532, "step": 173, "tokens/total": 5232800, "tokens/train_per_sec_per_gpu": 36.97, "tokens/trainable": 79059} +{"epoch": 0.6796875, "grad_norm": 0.19021031260490417, "learning_rate": 8.100214524900103e-05, "loss": 0.0018209717236459255, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.00182, "step": 174, "tokens/total": 5262672, "tokens/train_per_sec_per_gpu": 30.09, "tokens/trainable": 79498} +{"epoch": 0.68359375, "grad_norm": 0.03357760235667229, "learning_rate": 8.076468193435301e-05, "loss": 0.0004772062529809773, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00048, "step": 175, "tokens/total": 5293072, "tokens/train_per_sec_per_gpu": 30.82, "tokens/trainable": 79922} +{"epoch": 0.6875, "grad_norm": 0.41071707010269165, "learning_rate": 8.052614644612253e-05, "loss": 0.00923216249793768, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.46, "memory/max_allocated (GiB)": 33.46, "ppl": 1.00927, "step": 176, "tokens/total": 5321520, "tokens/train_per_sec_per_gpu": 36.09, "tokens/trainable": 80383} +{"epoch": 0.69140625, "grad_norm": 0.1126730740070343, "learning_rate": 8.028654871074489e-05, "loss": 0.0012899136636406183, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00129, "step": 177, "tokens/total": 5351904, "tokens/train_per_sec_per_gpu": 33.5, "tokens/trainable": 80824} +{"epoch": 0.6953125, "grad_norm": 0.06933529675006866, "learning_rate": 8.004589869885986e-05, "loss": 0.0005741716013289988, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00057, "step": 178, "tokens/total": 5382432, "tokens/train_per_sec_per_gpu": 32.7, "tokens/trainable": 81243} +{"epoch": 0.69921875, "grad_norm": 0.017655329778790474, "learning_rate": 7.980420642489674e-05, "loss": 0.0002967978944070637, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0003, "step": 179, "tokens/total": 5412960, "tokens/train_per_sec_per_gpu": 34.77, "tokens/trainable": 81716} +{"epoch": 0.703125, "grad_norm": 0.31013330817222595, "learning_rate": 7.95614819466576e-05, "loss": 0.0008732057176530361, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00087, "step": 180, "tokens/total": 5443232, "tokens/train_per_sec_per_gpu": 36.05, "tokens/trainable": 82181} +{"epoch": 0.70703125, "grad_norm": 0.4237620234489441, "learning_rate": 7.931773536489872e-05, "loss": 0.004198434762656689, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.29, "memory/max_allocated (GiB)": 33.29, "ppl": 1.00421, "step": 181, "tokens/total": 5471440, "tokens/train_per_sec_per_gpu": 35.13, "tokens/trainable": 82626} +{"epoch": 0.7109375, "grad_norm": 0.014504591003060341, "learning_rate": 7.907297682291035e-05, "loss": 0.00015536235878244042, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00016, "step": 182, "tokens/total": 5501744, "tokens/train_per_sec_per_gpu": 35.64, "tokens/trainable": 83089} +{"epoch": 0.71484375, "grad_norm": 0.0415632426738739, "learning_rate": 7.882721650609442e-05, "loss": 0.0004971292801201344, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0005, "step": 183, "tokens/total": 5532272, "tokens/train_per_sec_per_gpu": 35.84, "tokens/trainable": 83590} +{"epoch": 0.71875, "grad_norm": 0.16170728206634521, "learning_rate": 7.85804646415409e-05, "loss": 0.0023218044079840183, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00232, "step": 184, "tokens/total": 5562784, "tokens/train_per_sec_per_gpu": 35.77, "tokens/trainable": 84046} +{"epoch": 0.72265625, "grad_norm": 0.014852911233901978, "learning_rate": 7.833273149760207e-05, "loss": 0.0001844349899329245, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00018, "step": 185, "tokens/total": 5592800, "tokens/train_per_sec_per_gpu": 34.46, "tokens/trainable": 84517} +{"epoch": 0.7265625, "grad_norm": 0.4728831350803375, "learning_rate": 7.808402738346527e-05, "loss": 0.006746332161128521, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00677, "step": 186, "tokens/total": 5623088, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 84974} +{"epoch": 0.73046875, "grad_norm": 0.03623728081583977, "learning_rate": 7.783436264872382e-05, "loss": 0.0004019437765236944, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0004, "step": 187, "tokens/total": 5653344, "tokens/train_per_sec_per_gpu": 35.52, "tokens/trainable": 85460} +{"epoch": 0.734375, "grad_norm": 0.2675865888595581, "learning_rate": 7.758374768294647e-05, "loss": 0.0020086662843823433, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00201, "step": 188, "tokens/total": 5683600, "tokens/train_per_sec_per_gpu": 36.52, "tokens/trainable": 85939} +{"epoch": 0.73828125, "grad_norm": 0.507493793964386, "learning_rate": 7.733219291524489e-05, "loss": 0.0029892839957028627, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00299, "step": 189, "tokens/total": 5711952, "tokens/train_per_sec_per_gpu": 38.92, "tokens/trainable": 86377} +{"epoch": 0.7421875, "grad_norm": 0.0674358382821083, "learning_rate": 7.707970881383977e-05, "loss": 0.0005840241792611778, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00058, "step": 190, "tokens/total": 5742336, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 86834} +{"epoch": 0.74609375, "grad_norm": 1.451257348060608, "learning_rate": 7.682630588562518e-05, "loss": 0.010650807991623878, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01071, "step": 191, "tokens/total": 5772560, "tokens/train_per_sec_per_gpu": 32.48, "tokens/trainable": 87265} +{"epoch": 0.75, "grad_norm": 0.7005940079689026, "learning_rate": 7.657199467573129e-05, "loss": 0.012993288226425648, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01308, "step": 192, "tokens/total": 5803136, "tokens/train_per_sec_per_gpu": 34.56, "tokens/trainable": 87747} +{"epoch": 0.75390625, "grad_norm": 0.08145342022180557, "learning_rate": 7.631678576708561e-05, "loss": 0.0008674697601236403, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00087, "step": 193, "tokens/total": 5833440, "tokens/train_per_sec_per_gpu": 36.63, "tokens/trainable": 88239} +{"epoch": 0.7578125, "grad_norm": 0.07144517451524734, "learning_rate": 7.606068977997255e-05, "loss": 0.000845041242428124, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00085, "step": 194, "tokens/total": 5863552, "tokens/train_per_sec_per_gpu": 36.74, "tokens/trainable": 88718} +{"epoch": 0.76171875, "grad_norm": 0.4820076823234558, "learning_rate": 7.580371737159148e-05, "loss": 0.00316348928026855, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00317, "step": 195, "tokens/total": 5893680, "tokens/train_per_sec_per_gpu": 31.63, "tokens/trainable": 89129} +{"epoch": 0.765625, "grad_norm": 0.008089586161077023, "learning_rate": 7.554587923561324e-05, "loss": 0.00014078138337936252, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00014, "step": 196, "tokens/total": 5923744, "tokens/train_per_sec_per_gpu": 32.74, "tokens/trainable": 89567} +{"epoch": 0.76953125, "grad_norm": 0.4833170175552368, "learning_rate": 7.528718610173511e-05, "loss": 0.011923539452254772, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01199, "step": 197, "tokens/total": 5954128, "tokens/train_per_sec_per_gpu": 30.75, "tokens/trainable": 89988} +{"epoch": 0.7734375, "grad_norm": 0.4297773540019989, "learning_rate": 7.502764873523431e-05, "loss": 0.006487303879112005, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00651, "step": 198, "tokens/total": 5984352, "tokens/train_per_sec_per_gpu": 36.58, "tokens/trainable": 90434} +{"epoch": 0.77734375, "grad_norm": 0.30053746700286865, "learning_rate": 7.476727793652011e-05, "loss": 0.0037448552902787924, "memory/device_reserved (GiB)": 35.6, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00375, "step": 199, "tokens/total": 6014704, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 90913} +{"epoch": 0.78125, "grad_norm": 0.5079417824745178, "learning_rate": 7.450608454068415e-05, "loss": 0.01055203191936016, "memory/device_reserved (GiB)": 35.66, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.01061, "step": 200, "tokens/total": 6045056, "tokens/train_per_sec_per_gpu": 34.65, "tokens/trainable": 91355} +{"epoch": 0.78515625, "grad_norm": 0.3676057755947113, "learning_rate": 7.424407941704987e-05, "loss": 0.00716105941683054, "memory/device_reserved (GiB)": 35.66, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00719, "step": 201, "tokens/total": 6075504, "tokens/train_per_sec_per_gpu": 37.75, "tokens/trainable": 91856} +{"epoch": 0.7890625, "grad_norm": 0.06194091588258743, "learning_rate": 7.398127346871986e-05, "loss": 0.0004702930455096066, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00047, "step": 202, "tokens/total": 6105904, "tokens/train_per_sec_per_gpu": 34.56, "tokens/trainable": 92311} +{"epoch": 0.79296875, "grad_norm": 0.10208877176046371, "learning_rate": 7.371767763212238e-05, "loss": 0.0007544317631982267, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00075, "step": 203, "tokens/total": 6136272, "tokens/train_per_sec_per_gpu": 35.12, "tokens/trainable": 92764} +{"epoch": 0.796875, "grad_norm": 0.014593830332159996, "learning_rate": 7.345330287655617e-05, "loss": 0.00019093806622549891, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.65, "memory/max_allocated (GiB)": 33.65, "ppl": 1.00019, "step": 204, "tokens/total": 6166336, "tokens/train_per_sec_per_gpu": 33.48, "tokens/trainable": 93227} +{"epoch": 0.80078125, "grad_norm": 0.21324966847896576, "learning_rate": 7.31881602037339e-05, "loss": 0.0019866086076945066, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00199, "step": 205, "tokens/total": 6196720, "tokens/train_per_sec_per_gpu": 33.53, "tokens/trainable": 93675} +{"epoch": 0.8046875, "grad_norm": 0.012929766438901424, "learning_rate": 7.29222606473245e-05, "loss": 0.00013661546108778566, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00014, "step": 206, "tokens/total": 6227424, "tokens/train_per_sec_per_gpu": 32.77, "tokens/trainable": 94120} +{"epoch": 0.80859375, "grad_norm": 0.48922187089920044, "learning_rate": 7.265561527249383e-05, "loss": 0.014999452978372574, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.01511, "step": 207, "tokens/total": 6257616, "tokens/train_per_sec_per_gpu": 35.75, "tokens/trainable": 94557} +{"epoch": 0.8125, "grad_norm": 0.5681560039520264, "learning_rate": 7.238823517544436e-05, "loss": 0.011948533356189728, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01202, "step": 208, "tokens/total": 6288000, "tokens/train_per_sec_per_gpu": 41.01, "tokens/trainable": 95035} +{"epoch": 0.81640625, "grad_norm": 0.011107727885246277, "learning_rate": 7.212013148295333e-05, "loss": 0.00015826086746528745, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00016, "step": 209, "tokens/total": 6318336, "tokens/train_per_sec_per_gpu": 37.95, "tokens/trainable": 95517} +{"epoch": 0.8203125, "grad_norm": 0.044534292072057724, "learning_rate": 7.185131535190975e-05, "loss": 0.0009664896642789245, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00097, "step": 210, "tokens/total": 6348656, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 96026} +{"epoch": 0.82421875, "grad_norm": 0.7783606648445129, "learning_rate": 7.158179796885005e-05, "loss": 0.02202792465686798, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.02227, "step": 211, "tokens/total": 6379040, "tokens/train_per_sec_per_gpu": 35.83, "tokens/trainable": 96477} +{"epoch": 0.828125, "grad_norm": 0.17146360874176025, "learning_rate": 7.131159054949273e-05, "loss": 0.0028479855973273516, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00285, "step": 212, "tokens/total": 6409312, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 96951} +{"epoch": 0.83203125, "grad_norm": 0.19646775722503662, "learning_rate": 7.104070433827139e-05, "loss": 0.00397948594763875, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00399, "step": 213, "tokens/total": 6439520, "tokens/train_per_sec_per_gpu": 31.1, "tokens/trainable": 97350} +{"epoch": 0.8359375, "grad_norm": 0.06095995381474495, "learning_rate": 7.076915060786705e-05, "loss": 0.0011619608849287033, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00116, "step": 214, "tokens/total": 6469888, "tokens/train_per_sec_per_gpu": 30.13, "tokens/trainable": 97792} +{"epoch": 0.83984375, "grad_norm": 0.08135152608156204, "learning_rate": 7.049694065873882e-05, "loss": 0.0027059789281338453, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00271, "step": 215, "tokens/total": 6500128, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 98257} +{"epoch": 0.84375, "grad_norm": 0.14376476407051086, "learning_rate": 7.022408581865382e-05, "loss": 0.002945779822766781, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00295, "step": 216, "tokens/total": 6530832, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 98731} +{"epoch": 0.84765625, "grad_norm": 0.030300427228212357, "learning_rate": 6.99505974422157e-05, "loss": 0.0006251530721783638, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00063, "step": 217, "tokens/total": 6561248, "tokens/train_per_sec_per_gpu": 36.81, "tokens/trainable": 99212} +{"epoch": 0.8515625, "grad_norm": 0.08062481135129929, "learning_rate": 6.967648691039213e-05, "loss": 0.002811428625136614, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00282, "step": 218, "tokens/total": 6591648, "tokens/train_per_sec_per_gpu": 33.37, "tokens/trainable": 99654} +{"epoch": 0.85546875, "grad_norm": 0.21464644372463226, "learning_rate": 6.940176563004123e-05, "loss": 0.004250328987836838, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00426, "step": 219, "tokens/total": 6622368, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 100086} +{"epoch": 0.859375, "grad_norm": 0.06321574747562408, "learning_rate": 6.912644503343682e-05, "loss": 0.0014506221050396562, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00145, "step": 220, "tokens/total": 6652848, "tokens/train_per_sec_per_gpu": 33.58, "tokens/trainable": 100524} +{"epoch": 0.86328125, "grad_norm": 0.07537315040826797, "learning_rate": 6.885053657779273e-05, "loss": 0.0018983217887580395, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0019, "step": 221, "tokens/total": 6683392, "tokens/train_per_sec_per_gpu": 33.65, "tokens/trainable": 100996} +{"epoch": 0.8671875, "grad_norm": 0.04364407807588577, "learning_rate": 6.857405174478604e-05, "loss": 0.0008961158455349505, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0009, "step": 222, "tokens/total": 6713712, "tokens/train_per_sec_per_gpu": 32.28, "tokens/trainable": 101449} +{"epoch": 0.87109375, "grad_norm": 0.28126901388168335, "learning_rate": 6.82970020400792e-05, "loss": 0.002431782428175211, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00243, "step": 223, "tokens/total": 6744176, "tokens/train_per_sec_per_gpu": 33.36, "tokens/trainable": 101905} +{"epoch": 0.875, "grad_norm": 0.04008863866329193, "learning_rate": 6.801939899284132e-05, "loss": 0.0008201852906495333, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00082, "step": 224, "tokens/total": 6774416, "tokens/train_per_sec_per_gpu": 36.88, "tokens/trainable": 102411} +{"epoch": 0.87890625, "grad_norm": 0.09057830274105072, "learning_rate": 6.774125415526827e-05, "loss": 0.00176344346255064, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00176, "step": 225, "tokens/total": 6804592, "tokens/train_per_sec_per_gpu": 34.18, "tokens/trainable": 102882} +{"epoch": 0.8828125, "grad_norm": 0.40913277864456177, "learning_rate": 6.746257910210214e-05, "loss": 0.00800992175936699, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00804, "step": 226, "tokens/total": 6834976, "tokens/train_per_sec_per_gpu": 34.9, "tokens/trainable": 103332} +{"epoch": 0.88671875, "grad_norm": 0.06365697085857391, "learning_rate": 6.718338543014937e-05, "loss": 0.0008380536455661058, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00084, "step": 227, "tokens/total": 6865408, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 103790} +{"epoch": 0.890625, "grad_norm": 0.005372561980038881, "learning_rate": 6.69036847577983e-05, "loss": 8.878827793523669e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00009, "step": 228, "tokens/total": 6895664, "tokens/train_per_sec_per_gpu": 34.62, "tokens/trainable": 104246} +{"epoch": 0.89453125, "grad_norm": 0.3291122615337372, "learning_rate": 6.662348872453553e-05, "loss": 0.005115922540426254, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00513, "step": 229, "tokens/total": 6926016, "tokens/train_per_sec_per_gpu": 37.99, "tokens/trainable": 104707} +{"epoch": 0.8984375, "grad_norm": 0.005812515504658222, "learning_rate": 6.63428089904618e-05, "loss": 9.655929170548916e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0001, "step": 230, "tokens/total": 6956384, "tokens/train_per_sec_per_gpu": 37.24, "tokens/trainable": 105167} +{"epoch": 0.90234375, "grad_norm": 0.24782787263393402, "learning_rate": 6.60616572358065e-05, "loss": 0.002415842143818736, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00242, "step": 231, "tokens/total": 6986768, "tokens/train_per_sec_per_gpu": 33.5, "tokens/trainable": 105576} +{"epoch": 0.90625, "grad_norm": 0.04228518530726433, "learning_rate": 6.578004516044172e-05, "loss": 0.0006638517370447516, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00066, "step": 232, "tokens/total": 7017232, "tokens/train_per_sec_per_gpu": 37.24, "tokens/trainable": 106043} +{"epoch": 0.91015625, "grad_norm": 0.13585986196994781, "learning_rate": 6.549798448339548e-05, "loss": 0.0023324713110923767, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00234, "step": 233, "tokens/total": 7047568, "tokens/train_per_sec_per_gpu": 35.59, "tokens/trainable": 106506} +{"epoch": 0.9140625, "grad_norm": 0.13806359469890594, "learning_rate": 6.521548694236384e-05, "loss": 0.0027295195031911135, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00273, "step": 234, "tokens/total": 7077760, "tokens/train_per_sec_per_gpu": 34.5, "tokens/trainable": 106951} +{"epoch": 0.91796875, "grad_norm": 0.18551138043403625, "learning_rate": 6.493256429322259e-05, "loss": 0.0018342758994549513, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00184, "step": 235, "tokens/total": 7107760, "tokens/train_per_sec_per_gpu": 31.05, "tokens/trainable": 107382} +{"epoch": 0.921875, "grad_norm": 0.11544615775346756, "learning_rate": 6.464922830953799e-05, "loss": 0.002124304883182049, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00213, "step": 236, "tokens/total": 7138144, "tokens/train_per_sec_per_gpu": 33.23, "tokens/trainable": 107814} +{"epoch": 0.92578125, "grad_norm": 0.7617654800415039, "learning_rate": 6.436549078207688e-05, "loss": 0.014520522207021713, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01463, "step": 237, "tokens/total": 7168352, "tokens/train_per_sec_per_gpu": 35.7, "tokens/trainable": 108274} +{"epoch": 0.9296875, "grad_norm": 0.03847108036279678, "learning_rate": 6.408136351831592e-05, "loss": 0.0005749939009547234, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00058, "step": 238, "tokens/total": 7198624, "tokens/train_per_sec_per_gpu": 29.49, "tokens/trainable": 108714} +{"epoch": 0.93359375, "grad_norm": 0.13679052889347076, "learning_rate": 6.379685834195036e-05, "loss": 0.003290478838607669, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.0033, "step": 239, "tokens/total": 7229216, "tokens/train_per_sec_per_gpu": 37.14, "tokens/trainable": 109220} +{"epoch": 0.9375, "grad_norm": 0.1627172827720642, "learning_rate": 6.351198709240186e-05, "loss": 0.0017669250955805182, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00177, "step": 240, "tokens/total": 7259648, "tokens/train_per_sec_per_gpu": 32.42, "tokens/trainable": 109672} +{"epoch": 0.94140625, "grad_norm": 0.6043732166290283, "learning_rate": 6.32267616243259e-05, "loss": 0.015954118221998215, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01608, "step": 241, "tokens/total": 7289824, "tokens/train_per_sec_per_gpu": 35.51, "tokens/trainable": 110135} +{"epoch": 0.9453125, "grad_norm": 0.16619658470153809, "learning_rate": 6.294119380711849e-05, "loss": 0.0015951163368299603, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0016, "step": 242, "tokens/total": 7320288, "tokens/train_per_sec_per_gpu": 30.94, "tokens/trainable": 110563} +{"epoch": 0.94921875, "grad_norm": 0.024235516786575317, "learning_rate": 6.265529552442209e-05, "loss": 0.00045446690637618303, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00045, "step": 243, "tokens/total": 7350736, "tokens/train_per_sec_per_gpu": 37.45, "tokens/trainable": 111049} +{"epoch": 0.953125, "grad_norm": 0.10136919468641281, "learning_rate": 6.236907867363127e-05, "loss": 0.0024567395448684692, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00246, "step": 244, "tokens/total": 7381280, "tokens/train_per_sec_per_gpu": 33.71, "tokens/trainable": 111495} +{"epoch": 0.95703125, "grad_norm": 0.7829793691635132, "learning_rate": 6.208255516539749e-05, "loss": 0.004660450853407383, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00467, "step": 245, "tokens/total": 7411632, "tokens/train_per_sec_per_gpu": 32.12, "tokens/trainable": 111968} +{"epoch": 0.9609375, "grad_norm": 0.10806790739297867, "learning_rate": 6.179573692313344e-05, "loss": 0.001825700281187892, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00183, "step": 246, "tokens/total": 7441904, "tokens/train_per_sec_per_gpu": 36.87, "tokens/trainable": 112416} +{"epoch": 0.96484375, "grad_norm": 0.14372305572032928, "learning_rate": 6.150863588251694e-05, "loss": 0.0018829500768333673, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00188, "step": 247, "tokens/total": 7472368, "tokens/train_per_sec_per_gpu": 34.37, "tokens/trainable": 112882} +{"epoch": 0.96875, "grad_norm": 0.0468306727707386, "learning_rate": 6.122126399099419e-05, "loss": 0.001063495990820229, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00106, "step": 248, "tokens/total": 7502656, "tokens/train_per_sec_per_gpu": 40.56, "tokens/trainable": 113390} +{"epoch": 0.97265625, "grad_norm": 0.3564869165420532, "learning_rate": 6.0933633207282615e-05, "loss": 0.00227900268509984, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00228, "step": 249, "tokens/total": 7532800, "tokens/train_per_sec_per_gpu": 38.14, "tokens/trainable": 113904} +{"epoch": 0.9765625, "grad_norm": 0.02111588791012764, "learning_rate": 6.064575550087316e-05, "loss": 0.00027177060837857425, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00027, "step": 250, "tokens/total": 7563264, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 114351} +{"epoch": 0.98046875, "grad_norm": 0.2769322395324707, "learning_rate": 6.0357642851532245e-05, "loss": 0.004786302801221609, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.0048, "step": 251, "tokens/total": 7593840, "tokens/train_per_sec_per_gpu": 35.65, "tokens/trainable": 114842} +{"epoch": 0.984375, "grad_norm": 0.12309696525335312, "learning_rate": 6.0069307248803294e-05, "loss": 0.001197405974380672, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0012, "step": 252, "tokens/total": 7624416, "tokens/train_per_sec_per_gpu": 29.23, "tokens/trainable": 115263} +{"epoch": 0.98828125, "grad_norm": 0.6309447884559631, "learning_rate": 5.9780760691507635e-05, "loss": 0.009977569803595543, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01003, "step": 253, "tokens/total": 7654688, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 115703} +{"epoch": 0.9921875, "grad_norm": 0.022197438403964043, "learning_rate": 5.9492015187245334e-05, "loss": 0.0002314020530320704, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00023, "step": 254, "tokens/total": 7685088, "tokens/train_per_sec_per_gpu": 33.62, "tokens/trainable": 116157} +{"epoch": 0.99609375, "grad_norm": 0.13308732211589813, "learning_rate": 5.920308275189541e-05, "loss": 0.0008159374119713902, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00082, "step": 255, "tokens/total": 7715552, "tokens/train_per_sec_per_gpu": 32.39, "tokens/trainable": 116567} +{"epoch": 1.0, "grad_norm": 0.009844367392361164, "learning_rate": 5.8913975409115874e-05, "loss": 8.531998173566535e-05, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00009, "step": 256, "tokens/total": 7745872, "tokens/train_per_sec_per_gpu": 31.58, "tokens/trainable": 117030} +{"epoch": 1.00390625, "grad_norm": 0.008356324397027493, "learning_rate": 5.8624705189843395e-05, "loss": 8.252396219177172e-05, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00008, "step": 257, "tokens/total": 7776208, "tokens/train_per_sec_per_gpu": 35.13, "tokens/trainable": 117517} +{"epoch": 1.0078125, "grad_norm": 0.005978915374726057, "learning_rate": 5.833528413179249e-05, "loss": 6.622956425417215e-05, "memory/device_reserved (GiB)": 35.62, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00007, "step": 258, "tokens/total": 7806480, "tokens/train_per_sec_per_gpu": 38.82, "tokens/trainable": 118015} +{"epoch": 1.01171875, "grad_norm": 0.12491501867771149, "learning_rate": 5.80457242789548e-05, "loss": 0.0011766712414100766, "memory/device_reserved (GiB)": 35.62, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00118, "step": 259, "tokens/total": 7836800, "tokens/train_per_sec_per_gpu": 36.68, "tokens/trainable": 118520} +{"epoch": 1.015625, "grad_norm": 0.45353081822395325, "learning_rate": 5.77560376810977e-05, "loss": 0.029483629390597343, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02992, "step": 260, "tokens/total": 7867040, "tokens/train_per_sec_per_gpu": 36.67, "tokens/trainable": 118992} +{"epoch": 1.01953125, "grad_norm": 0.02936917543411255, "learning_rate": 5.7466236393263005e-05, "loss": 0.0003550515975803137, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00036, "step": 261, "tokens/total": 7897408, "tokens/train_per_sec_per_gpu": 34.24, "tokens/trainable": 119438} +{"epoch": 1.0234375, "grad_norm": 0.0044965120032429695, "learning_rate": 5.717633247526522e-05, "loss": 4.8715679440647364e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00005, "step": 262, "tokens/total": 7927648, "tokens/train_per_sec_per_gpu": 32.62, "tokens/trainable": 119881} +{"epoch": 1.02734375, "grad_norm": 0.11758256703615189, "learning_rate": 5.688633799118971e-05, "loss": 0.0004940429353155196, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00049, "step": 263, "tokens/total": 7957968, "tokens/train_per_sec_per_gpu": 28.22, "tokens/trainable": 120285} +{"epoch": 1.03125, "grad_norm": 0.003167251357808709, "learning_rate": 5.659626500889066e-05, "loss": 4.923924643662758e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.65, "memory/max_allocated (GiB)": 33.65, "ppl": 1.00005, "step": 264, "tokens/total": 7987888, "tokens/train_per_sec_per_gpu": 33.6, "tokens/trainable": 120755} +{"epoch": 1.03515625, "grad_norm": 0.0022706782910972834, "learning_rate": 5.6306125599488905e-05, "loss": 4.5954613597132266e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00005, "step": 265, "tokens/total": 8018048, "tokens/train_per_sec_per_gpu": 36.11, "tokens/trainable": 121194} +{"epoch": 1.0390625, "grad_norm": 0.2836518883705139, "learning_rate": 5.601593183686955e-05, "loss": 0.0011350277345627546, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00114, "step": 266, "tokens/total": 8048448, "tokens/train_per_sec_per_gpu": 37.14, "tokens/trainable": 121670} +{"epoch": 1.04296875, "grad_norm": 0.2555781304836273, "learning_rate": 5.572569579717961e-05, "loss": 0.0017866040579974651, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00179, "step": 267, "tokens/total": 8078848, "tokens/train_per_sec_per_gpu": 34.15, "tokens/trainable": 122142} +{"epoch": 1.046875, "grad_norm": 0.0076074316166341305, "learning_rate": 5.543542955832538e-05, "loss": 8.170487126335502e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00008, "step": 268, "tokens/total": 8109360, "tokens/train_per_sec_per_gpu": 32.86, "tokens/trainable": 122585} +{"epoch": 1.05078125, "grad_norm": 0.07716227322816849, "learning_rate": 5.514514519946986e-05, "loss": 0.0010607760632410645, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00106, "step": 269, "tokens/total": 8139888, "tokens/train_per_sec_per_gpu": 39.45, "tokens/trainable": 123091} +{"epoch": 1.0546875, "grad_norm": 0.018780827522277832, "learning_rate": 5.485485480053015e-05, "loss": 0.0002637105353642255, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00026, "step": 270, "tokens/total": 8170480, "tokens/train_per_sec_per_gpu": 31.42, "tokens/trainable": 123505} +{"epoch": 1.05859375, "grad_norm": 0.03091493807733059, "learning_rate": 5.4564570441674645e-05, "loss": 0.0003143992798868567, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.41, "memory/max_allocated (GiB)": 33.41, "ppl": 1.00031, "step": 271, "tokens/total": 8198848, "tokens/train_per_sec_per_gpu": 35.96, "tokens/trainable": 123953} +{"epoch": 1.0625, "grad_norm": 0.13643087446689606, "learning_rate": 5.42743042028204e-05, "loss": 0.0016409320523962379, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00164, "step": 272, "tokens/total": 8229200, "tokens/train_per_sec_per_gpu": 29.13, "tokens/trainable": 124400} +{"epoch": 1.06640625, "grad_norm": 0.15425634384155273, "learning_rate": 5.3984068163130464e-05, "loss": 0.0014877885114401579, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00149, "step": 273, "tokens/total": 8259536, "tokens/train_per_sec_per_gpu": 35.64, "tokens/trainable": 124870} +{"epoch": 1.0703125, "grad_norm": 0.12966743111610413, "learning_rate": 5.369387440051111e-05, "loss": 0.0022027853410691023, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00221, "step": 274, "tokens/total": 8289904, "tokens/train_per_sec_per_gpu": 32.54, "tokens/trainable": 125333} +{"epoch": 1.07421875, "grad_norm": 0.01629537157714367, "learning_rate": 5.340373499110935e-05, "loss": 0.00018065710901282728, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00018, "step": 275, "tokens/total": 8320176, "tokens/train_per_sec_per_gpu": 37.14, "tokens/trainable": 125834} +{"epoch": 1.078125, "grad_norm": 0.010340728797018528, "learning_rate": 5.3113662008810304e-05, "loss": 0.00013175717322155833, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00013, "step": 276, "tokens/total": 8350832, "tokens/train_per_sec_per_gpu": 38.6, "tokens/trainable": 126316} +{"epoch": 1.08203125, "grad_norm": 0.12547697126865387, "learning_rate": 5.282366752473479e-05, "loss": 0.0011799060739576817, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00118, "step": 277, "tokens/total": 8380976, "tokens/train_per_sec_per_gpu": 35.83, "tokens/trainable": 126798} +{"epoch": 1.0859375, "grad_norm": 0.013123616576194763, "learning_rate": 5.2533763606737005e-05, "loss": 8.206372876884416e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00008, "step": 278, "tokens/total": 8411072, "tokens/train_per_sec_per_gpu": 30.09, "tokens/trainable": 127223} +{"epoch": 1.08984375, "grad_norm": 0.02970624715089798, "learning_rate": 5.224396231890232e-05, "loss": 0.00019968026026617736, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.0002, "step": 279, "tokens/total": 8440736, "tokens/train_per_sec_per_gpu": 36.76, "tokens/trainable": 127672} +{"epoch": 1.09375, "grad_norm": 0.032595086842775345, "learning_rate": 5.195427572104522e-05, "loss": 0.0004402882477734238, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00044, "step": 280, "tokens/total": 8470944, "tokens/train_per_sec_per_gpu": 32.6, "tokens/trainable": 128116} +{"epoch": 1.09765625, "grad_norm": 0.007255643606185913, "learning_rate": 5.166471586820751e-05, "loss": 7.451212150044739e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00007, "step": 281, "tokens/total": 8501104, "tokens/train_per_sec_per_gpu": 33.24, "tokens/trainable": 128589} +{"epoch": 1.1015625, "grad_norm": 0.25917166471481323, "learning_rate": 5.1375294810156615e-05, "loss": 0.004500494804233313, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00451, "step": 282, "tokens/total": 8531696, "tokens/train_per_sec_per_gpu": 34.33, "tokens/trainable": 129036} +{"epoch": 1.10546875, "grad_norm": 0.016095004975795746, "learning_rate": 5.1086024590884144e-05, "loss": 0.00017428149294573814, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00017, "step": 283, "tokens/total": 8561936, "tokens/train_per_sec_per_gpu": 35.66, "tokens/trainable": 129485} +{"epoch": 1.109375, "grad_norm": 0.26327428221702576, "learning_rate": 5.079691724810461e-05, "loss": 0.006136145442724228, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00616, "step": 284, "tokens/total": 8592192, "tokens/train_per_sec_per_gpu": 29.96, "tokens/trainable": 129920} +{"epoch": 1.11328125, "grad_norm": 0.08115784078836441, "learning_rate": 5.0507984812754684e-05, "loss": 0.0008479508687742054, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00085, "step": 285, "tokens/total": 8622432, "tokens/train_per_sec_per_gpu": 36.61, "tokens/trainable": 130417} +{"epoch": 1.1171875, "grad_norm": 0.013995068147778511, "learning_rate": 5.021923930849237e-05, "loss": 0.00012496854469645768, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00012, "step": 286, "tokens/total": 8652768, "tokens/train_per_sec_per_gpu": 36.43, "tokens/trainable": 130903} +{"epoch": 1.12109375, "grad_norm": 0.005962018389254808, "learning_rate": 4.99306927511967e-05, "loss": 5.7330969866598025e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00006, "step": 287, "tokens/total": 8683296, "tokens/train_per_sec_per_gpu": 36.01, "tokens/trainable": 131343} +{"epoch": 1.125, "grad_norm": 0.005454273894429207, "learning_rate": 4.964235714846775e-05, "loss": 6.549475801875815e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00007, "step": 288, "tokens/total": 8713504, "tokens/train_per_sec_per_gpu": 32.63, "tokens/trainable": 131763} +{"epoch": 1.12890625, "grad_norm": 0.004979890305548906, "learning_rate": 4.9354244499126866e-05, "loss": 3.968320743297227e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00004, "step": 289, "tokens/total": 8743888, "tokens/train_per_sec_per_gpu": 34.48, "tokens/trainable": 132229} +{"epoch": 1.1328125, "grad_norm": 0.0025452408008277416, "learning_rate": 4.90663667927174e-05, "loss": 3.7232457543723285e-05, "memory/device_reserved (GiB)": 35.54, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00004, "step": 290, "tokens/total": 8774336, "tokens/train_per_sec_per_gpu": 35.87, "tokens/trainable": 132678} +{"epoch": 1.13671875, "grad_norm": 0.0009856978431344032, "learning_rate": 4.877873600900581e-05, "loss": 2.570721335359849e-05, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00003, "step": 291, "tokens/total": 8804816, "tokens/train_per_sec_per_gpu": 29.69, "tokens/trainable": 133136} +{"epoch": 1.140625, "grad_norm": 0.0006475941045209765, "learning_rate": 4.849136411748306e-05, "loss": 1.7373771697748452e-05, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00002, "step": 292, "tokens/total": 8835024, "tokens/train_per_sec_per_gpu": 35.73, "tokens/trainable": 133608} +{"epoch": 1.14453125, "grad_norm": 0.2577264606952667, "learning_rate": 4.8204263076866574e-05, "loss": 0.0018721828237175941, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00187, "step": 293, "tokens/total": 8865408, "tokens/train_per_sec_per_gpu": 40.93, "tokens/trainable": 134108} +{"epoch": 1.1484375, "grad_norm": 0.09887456148862839, "learning_rate": 4.791744483460251e-05, "loss": 0.0007890360429883003, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00079, "step": 294, "tokens/total": 8895824, "tokens/train_per_sec_per_gpu": 31.1, "tokens/trainable": 134541} +{"epoch": 1.15234375, "grad_norm": 0.005742037668824196, "learning_rate": 4.7630921326368736e-05, "loss": 5.973849329166114e-05, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00006, "step": 295, "tokens/total": 8926336, "tokens/train_per_sec_per_gpu": 29.83, "tokens/trainable": 134966} +{"epoch": 1.15625, "grad_norm": 0.0062770843505859375, "learning_rate": 4.7344704475577916e-05, "loss": 7.13710323907435e-05, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00007, "step": 296, "tokens/total": 8956656, "tokens/train_per_sec_per_gpu": 33.94, "tokens/trainable": 135402} +{"epoch": 1.16015625, "grad_norm": 0.009947159327566624, "learning_rate": 4.705880619288153e-05, "loss": 0.00010848060628632084, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00011, "step": 297, "tokens/total": 8986752, "tokens/train_per_sec_per_gpu": 30.41, "tokens/trainable": 135804} +{"epoch": 1.1640625, "grad_norm": 0.006360637955367565, "learning_rate": 4.677323837567412e-05, "loss": 5.382625022321008e-05, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00005, "step": 298, "tokens/total": 9017024, "tokens/train_per_sec_per_gpu": 31.54, "tokens/trainable": 136231} +{"epoch": 1.16796875, "grad_norm": 0.030510911718010902, "learning_rate": 4.6488012907598146e-05, "loss": 0.00030309066642075777, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0003, "step": 299, "tokens/total": 9047424, "tokens/train_per_sec_per_gpu": 34.68, "tokens/trainable": 136705} +{"epoch": 1.171875, "grad_norm": 0.004953032359480858, "learning_rate": 4.620314165804964e-05, "loss": 5.149709977558814e-05, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00005, "step": 300, "tokens/total": 9077648, "tokens/train_per_sec_per_gpu": 35.25, "tokens/trainable": 137178} +{"epoch": 1.17578125, "grad_norm": 0.003854556707665324, "learning_rate": 4.591863648168407e-05, "loss": 4.885891394224018e-05, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00005, "step": 301, "tokens/total": 9108032, "tokens/train_per_sec_per_gpu": 34.84, "tokens/trainable": 137643} +{"epoch": 1.1796875, "grad_norm": 0.11261430382728577, "learning_rate": 4.5634509217923135e-05, "loss": 0.0007564757252112031, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00076, "step": 302, "tokens/total": 9138240, "tokens/train_per_sec_per_gpu": 32.32, "tokens/trainable": 138078} +{"epoch": 1.18359375, "grad_norm": 0.03299031779170036, "learning_rate": 4.535077169046201e-05, "loss": 0.00041300299926660955, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00041, "step": 303, "tokens/total": 9168352, "tokens/train_per_sec_per_gpu": 35.5, "tokens/trainable": 138515} +{"epoch": 1.1875, "grad_norm": 0.0065305293537676334, "learning_rate": 4.506743570677743e-05, "loss": 7.41428229957819e-05, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00007, "step": 304, "tokens/total": 9198864, "tokens/train_per_sec_per_gpu": 35.62, "tokens/trainable": 138948} +{"epoch": 1.19140625, "grad_norm": 0.005171677563339472, "learning_rate": 4.478451305763618e-05, "loss": 6.253476749407127e-05, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00006, "step": 305, "tokens/total": 9229104, "tokens/train_per_sec_per_gpu": 38.57, "tokens/trainable": 139408} +{"epoch": 1.1953125, "grad_norm": 0.006580202840268612, "learning_rate": 4.450201551660454e-05, "loss": 5.272469570627436e-05, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.38, "memory/max_allocated (GiB)": 33.38, "ppl": 1.00005, "step": 306, "tokens/total": 9257344, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 139840} +{"epoch": 1.19921875, "grad_norm": 0.31961339712142944, "learning_rate": 4.4219954839558276e-05, "loss": 0.003208739683032036, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00321, "step": 307, "tokens/total": 9287520, "tokens/train_per_sec_per_gpu": 33.29, "tokens/trainable": 140281} +{"epoch": 1.203125, "grad_norm": 0.0023711167741566896, "learning_rate": 4.393834276419352e-05, "loss": 3.434348400332965e-05, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00003, "step": 308, "tokens/total": 9317888, "tokens/train_per_sec_per_gpu": 37.87, "tokens/trainable": 140776} +{"epoch": 1.20703125, "grad_norm": 0.019930923357605934, "learning_rate": 4.36571910095382e-05, "loss": 0.00012936958228237927, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00013, "step": 309, "tokens/total": 9348256, "tokens/train_per_sec_per_gpu": 37.31, "tokens/trainable": 141228} +{"epoch": 1.2109375, "grad_norm": 0.0023675146512687206, "learning_rate": 4.337651127546448e-05, "loss": 2.4802844563964754e-05, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00002, "step": 310, "tokens/total": 9378560, "tokens/train_per_sec_per_gpu": 37.82, "tokens/trainable": 141679} +{"epoch": 1.21484375, "grad_norm": 0.0036730507854372263, "learning_rate": 4.3096315242201736e-05, "loss": 3.144190486636944e-05, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00003, "step": 311, "tokens/total": 9408848, "tokens/train_per_sec_per_gpu": 32.72, "tokens/trainable": 142122} +{"epoch": 1.21875, "grad_norm": 0.0003704871633090079, "learning_rate": 4.2816614569850635e-05, "loss": 8.931780030252412e-06, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00001, "step": 312, "tokens/total": 9439344, "tokens/train_per_sec_per_gpu": 33.43, "tokens/trainable": 142561} +{"epoch": 1.22265625, "grad_norm": 0.05240330100059509, "learning_rate": 4.2537420897897864e-05, "loss": 0.00026115920627489686, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00026, "step": 313, "tokens/total": 9469792, "tokens/train_per_sec_per_gpu": 36.04, "tokens/trainable": 143046} +{"epoch": 1.2265625, "grad_norm": 0.0012482519960030913, "learning_rate": 4.225874584473174e-05, "loss": 2.5392351744812913e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 314, "tokens/total": 9500128, "tokens/train_per_sec_per_gpu": 33.39, "tokens/trainable": 143493} +{"epoch": 1.23046875, "grad_norm": 0.050520945340394974, "learning_rate": 4.19806010071587e-05, "loss": 0.0006884933682158589, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00069, "step": 315, "tokens/total": 9530480, "tokens/train_per_sec_per_gpu": 35.65, "tokens/trainable": 143956} +{"epoch": 1.234375, "grad_norm": 0.002656160155311227, "learning_rate": 4.170299795992081e-05, "loss": 2.599508843559306e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00003, "step": 316, "tokens/total": 9560912, "tokens/train_per_sec_per_gpu": 34.36, "tokens/trainable": 144411} +{"epoch": 1.23828125, "grad_norm": 0.19733326137065887, "learning_rate": 4.142594825521398e-05, "loss": 0.0016555915353819728, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00166, "step": 317, "tokens/total": 9591344, "tokens/train_per_sec_per_gpu": 39.15, "tokens/trainable": 144892} +{"epoch": 1.2421875, "grad_norm": 0.008795367553830147, "learning_rate": 4.114946342220728e-05, "loss": 6.713026959914714e-05, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00007, "step": 318, "tokens/total": 9621392, "tokens/train_per_sec_per_gpu": 33.07, "tokens/trainable": 145339} +{"epoch": 1.24609375, "grad_norm": 0.00049841299187392, "learning_rate": 4.087355496656321e-05, "loss": 1.1344342965458054e-05, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00001, "step": 319, "tokens/total": 9651728, "tokens/train_per_sec_per_gpu": 36.77, "tokens/trainable": 145811} +{"epoch": 1.25, "grad_norm": 0.23748824000358582, "learning_rate": 4.05982343699588e-05, "loss": 0.0030270391143858433, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00303, "step": 320, "tokens/total": 9682224, "tokens/train_per_sec_per_gpu": 39.19, "tokens/trainable": 146314} +{"epoch": 1.25390625, "grad_norm": 0.0024427378084510565, "learning_rate": 4.0323513089607876e-05, "loss": 3.495265627861954e-05, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00003, "step": 321, "tokens/total": 9712656, "tokens/train_per_sec_per_gpu": 32.1, "tokens/trainable": 146781} +{"epoch": 1.2578125, "grad_norm": 0.01809818297624588, "learning_rate": 4.004940255778431e-05, "loss": 0.00013907425454817712, "memory/device_reserved (GiB)": 34.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00014, "step": 322, "tokens/total": 9743088, "tokens/train_per_sec_per_gpu": 35.95, "tokens/trainable": 147225} +{"epoch": 1.26171875, "grad_norm": 0.0010443986393511295, "learning_rate": 3.977591418134619e-05, "loss": 1.951586818904616e-05, "memory/device_reserved (GiB)": 35.17, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00002, "step": 323, "tokens/total": 9773808, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 147673} +{"epoch": 1.265625, "grad_norm": 0.009523316286504269, "learning_rate": 3.95030593412612e-05, "loss": 6.634376040892676e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00007, "step": 324, "tokens/total": 9804016, "tokens/train_per_sec_per_gpu": 33.62, "tokens/trainable": 148103} +{"epoch": 1.26953125, "grad_norm": 0.049099843949079514, "learning_rate": 3.923084939213296e-05, "loss": 0.00039430244942195714, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00039, "step": 325, "tokens/total": 9834592, "tokens/train_per_sec_per_gpu": 32.11, "tokens/trainable": 148535} +{"epoch": 1.2734375, "grad_norm": 0.01699669286608696, "learning_rate": 3.895929566172861e-05, "loss": 0.00013168745499569923, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00013, "step": 326, "tokens/total": 9864848, "tokens/train_per_sec_per_gpu": 34.55, "tokens/trainable": 148999} +{"epoch": 1.27734375, "grad_norm": 0.07120174169540405, "learning_rate": 3.868840945050728e-05, "loss": 0.000350497430190444, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00035, "step": 327, "tokens/total": 9895168, "tokens/train_per_sec_per_gpu": 32.05, "tokens/trainable": 149431} +{"epoch": 1.28125, "grad_norm": 0.001589596620760858, "learning_rate": 3.841820203114995e-05, "loss": 2.160290750907734e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00002, "step": 328, "tokens/total": 9925696, "tokens/train_per_sec_per_gpu": 34.56, "tokens/trainable": 149886} +{"epoch": 1.28515625, "grad_norm": 0.014652811922132969, "learning_rate": 3.814868464809027e-05, "loss": 0.00011893574264831841, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00012, "step": 329, "tokens/total": 9955648, "tokens/train_per_sec_per_gpu": 31.44, "tokens/trainable": 150288} +{"epoch": 1.2890625, "grad_norm": 0.032810863107442856, "learning_rate": 3.787986851704667e-05, "loss": 0.00019718095427379012, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0002, "step": 330, "tokens/total": 9985856, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 150733} +{"epoch": 1.29296875, "grad_norm": 0.02765781246125698, "learning_rate": 3.7611764824555654e-05, "loss": 0.00016869997489266098, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00017, "step": 331, "tokens/total": 10016208, "tokens/train_per_sec_per_gpu": 37.27, "tokens/trainable": 151207} +{"epoch": 1.296875, "grad_norm": 0.01766805537045002, "learning_rate": 3.734438472750619e-05, "loss": 0.00014554498193319887, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00015, "step": 332, "tokens/total": 10046512, "tokens/train_per_sec_per_gpu": 38.07, "tokens/trainable": 151694} +{"epoch": 1.30078125, "grad_norm": 0.05623512342572212, "learning_rate": 3.707773935267552e-05, "loss": 0.00024310205481015146, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00024, "step": 333, "tokens/total": 10076944, "tokens/train_per_sec_per_gpu": 38.84, "tokens/trainable": 152188} +{"epoch": 1.3046875, "grad_norm": 0.0034802353475242853, "learning_rate": 3.68118397962661e-05, "loss": 3.652479063021019e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00004, "step": 334, "tokens/total": 10107296, "tokens/train_per_sec_per_gpu": 31.03, "tokens/trainable": 152596} +{"epoch": 1.30859375, "grad_norm": 0.0005486137815751135, "learning_rate": 3.654669712344384e-05, "loss": 1.2760298886860255e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00001, "step": 335, "tokens/total": 10137568, "tokens/train_per_sec_per_gpu": 36.99, "tokens/trainable": 153052} +{"epoch": 1.3125, "grad_norm": 0.053988587111234665, "learning_rate": 3.628232236787763e-05, "loss": 0.0004493523156270385, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00045, "step": 336, "tokens/total": 10167952, "tokens/train_per_sec_per_gpu": 30.61, "tokens/trainable": 153491} +{"epoch": 1.31640625, "grad_norm": 0.023256463930010796, "learning_rate": 3.6018726531280144e-05, "loss": 0.0001921930379467085, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00019, "step": 337, "tokens/total": 10198512, "tokens/train_per_sec_per_gpu": 40.87, "tokens/trainable": 153983} +{"epoch": 1.3203125, "grad_norm": 0.005770122166723013, "learning_rate": 3.575592058295017e-05, "loss": 5.241332473815419e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00005, "step": 338, "tokens/total": 10228752, "tokens/train_per_sec_per_gpu": 37.15, "tokens/trainable": 154464} +{"epoch": 1.32421875, "grad_norm": 0.0003110536199528724, "learning_rate": 3.549391545931585e-05, "loss": 5.897365099372109e-06, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00001, "step": 339, "tokens/total": 10259104, "tokens/train_per_sec_per_gpu": 36.25, "tokens/trainable": 154924} +{"epoch": 1.328125, "grad_norm": 0.008365475572645664, "learning_rate": 3.5232722063479914e-05, "loss": 7.736510451650247e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00008, "step": 340, "tokens/total": 10289520, "tokens/train_per_sec_per_gpu": 31.2, "tokens/trainable": 155373} +{"epoch": 1.33203125, "grad_norm": 0.003012418746948242, "learning_rate": 3.49723512647657e-05, "loss": 3.9521881262771785e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00004, "step": 341, "tokens/total": 10320016, "tokens/train_per_sec_per_gpu": 36.92, "tokens/trainable": 155873} +{"epoch": 1.3359375, "grad_norm": 0.00036992691457271576, "learning_rate": 3.471281389826491e-05, "loss": 9.513973964203615e-06, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00001, "step": 342, "tokens/total": 10350368, "tokens/train_per_sec_per_gpu": 30.77, "tokens/trainable": 156278} +{"epoch": 1.33984375, "grad_norm": 0.004851094912737608, "learning_rate": 3.4454120764386764e-05, "loss": 3.941698741982691e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00004, "step": 343, "tokens/total": 10380624, "tokens/train_per_sec_per_gpu": 35.8, "tokens/trainable": 156733} +{"epoch": 1.34375, "grad_norm": 0.08602030575275421, "learning_rate": 3.4196282628408526e-05, "loss": 0.0006739848759025335, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00067, "step": 344, "tokens/total": 10411024, "tokens/train_per_sec_per_gpu": 34.27, "tokens/trainable": 157203} +{"epoch": 1.34765625, "grad_norm": 0.007851852104067802, "learning_rate": 3.3939310220027456e-05, "loss": 7.82167335273698e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00008, "step": 345, "tokens/total": 10441248, "tokens/train_per_sec_per_gpu": 38.34, "tokens/trainable": 157707} +{"epoch": 1.3515625, "grad_norm": 0.000522423128131777, "learning_rate": 3.3683214232914404e-05, "loss": 7.171449851739453e-06, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00001, "step": 346, "tokens/total": 10471712, "tokens/train_per_sec_per_gpu": 35.37, "tokens/trainable": 158180} +{"epoch": 1.35546875, "grad_norm": 0.0008156000403687358, "learning_rate": 3.342800532426873e-05, "loss": 1.2959781088284217e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00001, "step": 347, "tokens/total": 10502288, "tokens/train_per_sec_per_gpu": 34.8, "tokens/trainable": 158646} +{"epoch": 1.359375, "grad_norm": 0.012755601666867733, "learning_rate": 3.317369411437484e-05, "loss": 0.00010119444777956232, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0001, "step": 348, "tokens/total": 10532304, "tokens/train_per_sec_per_gpu": 36.12, "tokens/trainable": 159115} +{"epoch": 1.36328125, "grad_norm": 0.0019102037185803056, "learning_rate": 3.292029118616024e-05, "loss": 2.8528616894618608e-05, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00003, "step": 349, "tokens/total": 10562608, "tokens/train_per_sec_per_gpu": 31.94, "tokens/trainable": 159538} +{"epoch": 1.3671875, "grad_norm": 0.07546474039554596, "learning_rate": 3.266780708475511e-05, "loss": 0.0005027748411521316, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0005, "step": 350, "tokens/total": 10592992, "tokens/train_per_sec_per_gpu": 33.6, "tokens/trainable": 160016} +{"epoch": 1.37109375, "grad_norm": 0.014789975248277187, "learning_rate": 3.241625231705354e-05, "loss": 0.00011153960804222152, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00011, "step": 351, "tokens/total": 10623328, "tokens/train_per_sec_per_gpu": 34.93, "tokens/trainable": 160475} +{"epoch": 1.375, "grad_norm": 0.001108153141103685, "learning_rate": 3.216563735127618e-05, "loss": 1.160802275990136e-05, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00001, "step": 352, "tokens/total": 10653632, "tokens/train_per_sec_per_gpu": 34.37, "tokens/trainable": 160952} +{"epoch": 1.37890625, "grad_norm": 0.061208564788103104, "learning_rate": 3.191597261653475e-05, "loss": 0.0005815322510898113, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00058, "step": 353, "tokens/total": 10684176, "tokens/train_per_sec_per_gpu": 33.27, "tokens/trainable": 161420} +{"epoch": 1.3828125, "grad_norm": 0.0031761936843395233, "learning_rate": 3.166726850239794e-05, "loss": 2.453591878293082e-05, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00002, "step": 354, "tokens/total": 10714352, "tokens/train_per_sec_per_gpu": 34.49, "tokens/trainable": 161868} +{"epoch": 1.38671875, "grad_norm": 0.0003757224476430565, "learning_rate": 3.141953535845912e-05, "loss": 6.875088274682639e-06, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00001, "step": 355, "tokens/total": 10744464, "tokens/train_per_sec_per_gpu": 35.79, "tokens/trainable": 162318} +{"epoch": 1.390625, "grad_norm": 0.0009734284249134362, "learning_rate": 3.11727834939056e-05, "loss": 9.81441462499788e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00001, "step": 356, "tokens/total": 10774976, "tokens/train_per_sec_per_gpu": 31.85, "tokens/trainable": 162789} +{"epoch": 1.39453125, "grad_norm": 0.0010252115316689014, "learning_rate": 3.092702317708967e-05, "loss": 8.114137017400935e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00001, "step": 357, "tokens/total": 10805600, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 163254} +{"epoch": 1.3984375, "grad_norm": 0.0002301523054484278, "learning_rate": 3.0682264635101276e-05, "loss": 5.338565642887261e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00001, "step": 358, "tokens/total": 10835920, "tokens/train_per_sec_per_gpu": 34.92, "tokens/trainable": 163715} +{"epoch": 1.40234375, "grad_norm": 0.0006164724472910166, "learning_rate": 3.0438518053342407e-05, "loss": 8.090713890851475e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00001, "step": 359, "tokens/total": 10866048, "tokens/train_per_sec_per_gpu": 34.88, "tokens/trainable": 164197} +{"epoch": 1.40625, "grad_norm": 0.0006751597975380719, "learning_rate": 3.0195793575103266e-05, "loss": 1.1418029316700995e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00001, "step": 360, "tokens/total": 10896288, "tokens/train_per_sec_per_gpu": 32.03, "tokens/trainable": 164661} +{"epoch": 1.41015625, "grad_norm": 0.00013061391655355692, "learning_rate": 2.9954101301140146e-05, "loss": 3.586079174056067e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0, "step": 361, "tokens/total": 10926816, "tokens/train_per_sec_per_gpu": 33.48, "tokens/trainable": 165111} +{"epoch": 1.4140625, "grad_norm": 0.001630443730391562, "learning_rate": 2.9713451289255123e-05, "loss": 1.7084981664083898e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.31, "memory/max_allocated (GiB)": 33.31, "ppl": 1.00002, "step": 362, "tokens/total": 10954928, "tokens/train_per_sec_per_gpu": 30.82, "tokens/trainable": 165552} +{"epoch": 1.41796875, "grad_norm": 0.0003005889884661883, "learning_rate": 2.9473853553877484e-05, "loss": 6.793928150727879e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00001, "step": 363, "tokens/total": 10985328, "tokens/train_per_sec_per_gpu": 35.67, "tokens/trainable": 166026} +{"epoch": 1.421875, "grad_norm": 0.005686561111360788, "learning_rate": 2.9235318065647e-05, "loss": 5.100644557387568e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00005, "step": 364, "tokens/total": 11015664, "tokens/train_per_sec_per_gpu": 38.43, "tokens/trainable": 166486} +{"epoch": 1.42578125, "grad_norm": 0.0014047786826267838, "learning_rate": 2.8997854750998964e-05, "loss": 9.773429155757185e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00001, "step": 365, "tokens/total": 11046256, "tokens/train_per_sec_per_gpu": 36.88, "tokens/trainable": 166959} +{"epoch": 1.4296875, "grad_norm": 9.058098657988012e-05, "learning_rate": 2.8761473491751258e-05, "loss": 2.382128286626539e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.68, "memory/max_allocated (GiB)": 33.68, "ppl": 1.0, "step": 366, "tokens/total": 11076288, "tokens/train_per_sec_per_gpu": 34.15, "tokens/trainable": 167394} +{"epoch": 1.43359375, "grad_norm": 0.0009709852165542543, "learning_rate": 2.8526184124692883e-05, "loss": 7.577294582006289e-06, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00001, "step": 367, "tokens/total": 11106416, "tokens/train_per_sec_per_gpu": 28.89, "tokens/trainable": 167832} +{"epoch": 1.4375, "grad_norm": 0.0034374946262687445, "learning_rate": 2.829199644117484e-05, "loss": 3.723455301951617e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00004, "step": 368, "tokens/total": 11136848, "tokens/train_per_sec_per_gpu": 31.31, "tokens/trainable": 168242} +{"epoch": 1.44140625, "grad_norm": 0.0033002276904881, "learning_rate": 2.8058920186702553e-05, "loss": 2.274679354741238e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00002, "step": 369, "tokens/total": 11167376, "tokens/train_per_sec_per_gpu": 36.15, "tokens/trainable": 168715} +{"epoch": 1.4453125, "grad_norm": 0.0006988391978666186, "learning_rate": 2.782696506053033e-05, "loss": 5.0486655709391925e-06, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00001, "step": 370, "tokens/total": 11197776, "tokens/train_per_sec_per_gpu": 32.21, "tokens/trainable": 169147} +{"epoch": 1.44921875, "grad_norm": 0.0052696517668664455, "learning_rate": 2.7596140715257824e-05, "loss": 2.7391106414142996e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00003, "step": 371, "tokens/total": 11227840, "tokens/train_per_sec_per_gpu": 34.71, "tokens/trainable": 169628} +{"epoch": 1.453125, "grad_norm": 0.0015628528781235218, "learning_rate": 2.7366456756428184e-05, "loss": 1.4505098079098389e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00001, "step": 372, "tokens/total": 11258176, "tokens/train_per_sec_per_gpu": 35.59, "tokens/trainable": 170085} +{"epoch": 1.45703125, "grad_norm": 0.0021482163574546576, "learning_rate": 2.7137922742128486e-05, "loss": 1.399279062752612e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00001, "step": 373, "tokens/total": 11288336, "tokens/train_per_sec_per_gpu": 36.06, "tokens/trainable": 170584} +{"epoch": 1.4609375, "grad_norm": 0.001900387927889824, "learning_rate": 2.691054818259188e-05, "loss": 1.4173308954923414e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00001, "step": 374, "tokens/total": 11318848, "tokens/train_per_sec_per_gpu": 35.04, "tokens/trainable": 171058} +{"epoch": 1.46484375, "grad_norm": 0.003817511023953557, "learning_rate": 2.6684342539801933e-05, "loss": 2.5515182642266154e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00003, "step": 375, "tokens/total": 11349168, "tokens/train_per_sec_per_gpu": 34.37, "tokens/trainable": 171518} +{"epoch": 1.46875, "grad_norm": 0.3609258234500885, "learning_rate": 2.645931522709877e-05, "loss": 0.006828296929597855, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00685, "step": 376, "tokens/total": 11379328, "tokens/train_per_sec_per_gpu": 41.02, "tokens/trainable": 172007} +{"epoch": 1.47265625, "grad_norm": 0.0038110739551484585, "learning_rate": 2.6235475608787365e-05, "loss": 3.2214695238508284e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00003, "step": 377, "tokens/total": 11409568, "tokens/train_per_sec_per_gpu": 35.2, "tokens/trainable": 172468} +{"epoch": 1.4765625, "grad_norm": 0.0038299988955259323, "learning_rate": 2.6012832999747916e-05, "loss": 1.8610287952469662e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00002, "step": 378, "tokens/total": 11439968, "tokens/train_per_sec_per_gpu": 34.1, "tokens/trainable": 172923} +{"epoch": 1.48046875, "grad_norm": 0.24727243185043335, "learning_rate": 2.579139666504821e-05, "loss": 0.029269637539982796, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0297, "step": 379, "tokens/total": 11470496, "tokens/train_per_sec_per_gpu": 34.55, "tokens/trainable": 173370} +{"epoch": 1.484375, "grad_norm": 0.08204144984483719, "learning_rate": 2.557117581955798e-05, "loss": 0.0006305916467681527, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00063, "step": 380, "tokens/total": 11500720, "tokens/train_per_sec_per_gpu": 34.8, "tokens/trainable": 173837} +{"epoch": 1.48828125, "grad_norm": 0.016210485249757767, "learning_rate": 2.5352179627565532e-05, "loss": 9.964105993276462e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.0001, "step": 381, "tokens/total": 11531280, "tokens/train_per_sec_per_gpu": 35.99, "tokens/trainable": 174294} +{"epoch": 1.4921875, "grad_norm": 0.00014741058112122118, "learning_rate": 2.5134417202396277e-05, "loss": 4.461833668756299e-06, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0, "step": 382, "tokens/total": 11561264, "tokens/train_per_sec_per_gpu": 31.16, "tokens/trainable": 174701} +{"epoch": 1.49609375, "grad_norm": 0.00043298961827531457, "learning_rate": 2.491789760603361e-05, "loss": 1.0579917216091417e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00001, "step": 383, "tokens/total": 11591616, "tokens/train_per_sec_per_gpu": 32.49, "tokens/trainable": 175131} +{"epoch": 1.5, "grad_norm": 0.0009353766799904406, "learning_rate": 2.4702629848741764e-05, "loss": 1.3036698874202557e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00001, "step": 384, "tokens/total": 11622000, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 175586} +{"epoch": 1.50390625, "grad_norm": 0.004666164051741362, "learning_rate": 2.4488622888690785e-05, "loss": 5.6815359130268916e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00006, "step": 385, "tokens/total": 11652352, "tokens/train_per_sec_per_gpu": 34.36, "tokens/trainable": 176026} +{"epoch": 1.5078125, "grad_norm": 0.0007844759384170175, "learning_rate": 2.427588563158384e-05, "loss": 1.926398545037955e-05, "memory/device_reserved (GiB)": 34.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 386, "tokens/total": 11682736, "tokens/train_per_sec_per_gpu": 34.28, "tokens/trainable": 176512} +{"epoch": 1.51171875, "grad_norm": 0.0016678960528224707, "learning_rate": 2.406442693028651e-05, "loss": 2.7389756724005565e-05, "memory/device_reserved (GiB)": 34.98, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00003, "step": 387, "tokens/total": 11713136, "tokens/train_per_sec_per_gpu": 30.82, "tokens/trainable": 176943} +{"epoch": 1.515625, "grad_norm": 0.00107457279227674, "learning_rate": 2.3854255584458547e-05, "loss": 2.128102642018348e-05, "memory/device_reserved (GiB)": 35.23, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00002, "step": 388, "tokens/total": 11743808, "tokens/train_per_sec_per_gpu": 32.42, "tokens/trainable": 177370} +{"epoch": 1.51953125, "grad_norm": 0.0008889543823897839, "learning_rate": 2.3645380340187508e-05, "loss": 1.6756177501520142e-05, "memory/device_reserved (GiB)": 35.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00002, "step": 389, "tokens/total": 11774144, "tokens/train_per_sec_per_gpu": 35.56, "tokens/trainable": 177851} +{"epoch": 1.5234375, "grad_norm": 0.09012716263532639, "learning_rate": 2.3437809889624914e-05, "loss": 0.0013020217884331942, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.0013, "step": 390, "tokens/total": 11804272, "tokens/train_per_sec_per_gpu": 35.15, "tokens/trainable": 178310} +{"epoch": 1.52734375, "grad_norm": 0.02159537374973297, "learning_rate": 2.3231552870624487e-05, "loss": 0.00023354997392743826, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00023, "step": 391, "tokens/total": 11832432, "tokens/train_per_sec_per_gpu": 37.26, "tokens/trainable": 178736} +{"epoch": 1.53125, "grad_norm": 0.015429600141942501, "learning_rate": 2.3026617866382657e-05, "loss": 0.00018487832858227193, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00018, "step": 392, "tokens/total": 11862960, "tokens/train_per_sec_per_gpu": 36.68, "tokens/trainable": 179233} +{"epoch": 1.53515625, "grad_norm": 0.004461898934096098, "learning_rate": 2.2823013405081507e-05, "loss": 4.71106031909585e-05, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00005, "step": 393, "tokens/total": 11893232, "tokens/train_per_sec_per_gpu": 39.0, "tokens/trainable": 179730} +{"epoch": 1.5390625, "grad_norm": 0.0018961423775181174, "learning_rate": 2.2620747959533722e-05, "loss": 4.23475939896889e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00004, "step": 394, "tokens/total": 11923664, "tokens/train_per_sec_per_gpu": 38.28, "tokens/trainable": 180227} +{"epoch": 1.54296875, "grad_norm": 0.0012804885627701879, "learning_rate": 2.2419829946830123e-05, "loss": 2.206004865001887e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00002, "step": 395, "tokens/total": 11954032, "tokens/train_per_sec_per_gpu": 33.32, "tokens/trainable": 180687} +{"epoch": 1.546875, "grad_norm": 0.0014976069796830416, "learning_rate": 2.2220267727989325e-05, "loss": 2.556692015787121e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00003, "step": 396, "tokens/total": 11984512, "tokens/train_per_sec_per_gpu": 33.3, "tokens/trainable": 181141} +{"epoch": 1.55078125, "grad_norm": 0.005655917339026928, "learning_rate": 2.202206960760984e-05, "loss": 5.6497796322219074e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00006, "step": 397, "tokens/total": 12014880, "tokens/train_per_sec_per_gpu": 36.43, "tokens/trainable": 181648} +{"epoch": 1.5546875, "grad_norm": 0.01994832046329975, "learning_rate": 2.182524383352446e-05, "loss": 0.00019567189156077802, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.0002, "step": 398, "tokens/total": 12044928, "tokens/train_per_sec_per_gpu": 34.34, "tokens/trainable": 182109} +{"epoch": 1.55859375, "grad_norm": 0.004994136746972799, "learning_rate": 2.1629798596457056e-05, "loss": 4.0259299566969275e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00004, "step": 399, "tokens/total": 12075344, "tokens/train_per_sec_per_gpu": 31.99, "tokens/trainable": 182559} +{"epoch": 1.5625, "grad_norm": 0.0004713181406259537, "learning_rate": 2.1435742029681725e-05, "loss": 1.3051702808297705e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00001, "step": 400, "tokens/total": 12105616, "tokens/train_per_sec_per_gpu": 33.69, "tokens/trainable": 182985} +{"epoch": 1.56640625, "grad_norm": 0.00034305796725675464, "learning_rate": 2.124308220868431e-05, "loss": 1.262360456166789e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00001, "step": 401, "tokens/total": 12136080, "tokens/train_per_sec_per_gpu": 33.13, "tokens/trainable": 183440} +{"epoch": 1.5703125, "grad_norm": 0.15386679768562317, "learning_rate": 2.105182715082638e-05, "loss": 0.0018648281693458557, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00187, "step": 402, "tokens/total": 12166240, "tokens/train_per_sec_per_gpu": 34.96, "tokens/trainable": 183917} +{"epoch": 1.57421875, "grad_norm": 0.003915456123650074, "learning_rate": 2.0861984815011552e-05, "loss": 4.141925091971643e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00004, "step": 403, "tokens/total": 12196544, "tokens/train_per_sec_per_gpu": 30.7, "tokens/trainable": 184348} +{"epoch": 1.578125, "grad_norm": 0.0021876280661672354, "learning_rate": 2.0673563101354323e-05, "loss": 2.675295399967581e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00003, "step": 404, "tokens/total": 12226784, "tokens/train_per_sec_per_gpu": 31.9, "tokens/trainable": 184786} +{"epoch": 1.58203125, "grad_norm": 0.0006817729445174336, "learning_rate": 2.0486569850851317e-05, "loss": 1.70326584338909e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00002, "step": 405, "tokens/total": 12257264, "tokens/train_per_sec_per_gpu": 34.25, "tokens/trainable": 185249} +{"epoch": 1.5859375, "grad_norm": 0.0010018590837717056, "learning_rate": 2.0301012845054956e-05, "loss": 2.3183692974271253e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 406, "tokens/total": 12287584, "tokens/train_per_sec_per_gpu": 31.89, "tokens/trainable": 185680} +{"epoch": 1.58984375, "grad_norm": 0.001717094099149108, "learning_rate": 2.011689980574966e-05, "loss": 1.955918014573399e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00002, "step": 407, "tokens/total": 12317984, "tokens/train_per_sec_per_gpu": 32.08, "tokens/trainable": 186100} +{"epoch": 1.59375, "grad_norm": 0.0029396964237093925, "learning_rate": 1.993423839463052e-05, "loss": 3.9509199268650264e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00004, "step": 408, "tokens/total": 12348144, "tokens/train_per_sec_per_gpu": 40.3, "tokens/trainable": 186565} +{"epoch": 1.59765625, "grad_norm": 0.0014759147306904197, "learning_rate": 1.975303621298445e-05, "loss": 1.9150986190652475e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 409, "tokens/total": 12378544, "tokens/train_per_sec_per_gpu": 36.25, "tokens/trainable": 187017} +{"epoch": 1.6015625, "grad_norm": 0.002330441027879715, "learning_rate": 1.957330080137385e-05, "loss": 3.6902769352309406e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00004, "step": 410, "tokens/total": 12408784, "tokens/train_per_sec_per_gpu": 33.44, "tokens/trainable": 187468} +{"epoch": 1.60546875, "grad_norm": 0.0019221253460273147, "learning_rate": 1.9395039639322864e-05, "loss": 3.196115358150564e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00003, "step": 411, "tokens/total": 12438832, "tokens/train_per_sec_per_gpu": 38.07, "tokens/trainable": 187961} +{"epoch": 1.609375, "grad_norm": 0.17172899842262268, "learning_rate": 1.9218260145006073e-05, "loss": 0.0011913730995729566, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00119, "step": 412, "tokens/total": 12469296, "tokens/train_per_sec_per_gpu": 36.4, "tokens/trainable": 188447} +{"epoch": 1.61328125, "grad_norm": 0.006395282223820686, "learning_rate": 1.904296967493982e-05, "loss": 8.003453694982454e-05, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00008, "step": 413, "tokens/total": 12499392, "tokens/train_per_sec_per_gpu": 36.39, "tokens/trainable": 188903} +{"epoch": 1.6171875, "grad_norm": 0.03568729758262634, "learning_rate": 1.8869175523676064e-05, "loss": 0.0003129084361717105, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 34.0, "memory/max_allocated (GiB)": 34.0, "ppl": 1.00031, "step": 414, "tokens/total": 12529952, "tokens/train_per_sec_per_gpu": 37.08, "tokens/trainable": 189398} +{"epoch": 1.62109375, "grad_norm": 0.0005958312540315092, "learning_rate": 1.869688492349885e-05, "loss": 1.7803757145884447e-05, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00002, "step": 415, "tokens/total": 12560240, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 189850} +{"epoch": 1.625, "grad_norm": 0.04430711269378662, "learning_rate": 1.85261050441233e-05, "loss": 0.0004007165553048253, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0004, "step": 416, "tokens/total": 12590736, "tokens/train_per_sec_per_gpu": 33.12, "tokens/trainable": 190306} +{"epoch": 1.62890625, "grad_norm": 0.0005576324183493853, "learning_rate": 1.8356842992397304e-05, "loss": 1.260453700524522e-05, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00001, "step": 417, "tokens/total": 12621168, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 190746} +{"epoch": 1.6328125, "grad_norm": 0.0026761293411254883, "learning_rate": 1.8189105812005714e-05, "loss": 5.3298615966923535e-05, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00005, "step": 418, "tokens/total": 12651456, "tokens/train_per_sec_per_gpu": 31.53, "tokens/trainable": 191156} +{"epoch": 1.63671875, "grad_norm": 0.03599291294813156, "learning_rate": 1.802290048317732e-05, "loss": 0.0003095666761510074, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00031, "step": 419, "tokens/total": 12681760, "tokens/train_per_sec_per_gpu": 29.88, "tokens/trainable": 191573} +{"epoch": 1.640625, "grad_norm": 0.0009753056219778955, "learning_rate": 1.785823392239424e-05, "loss": 2.4977898647193797e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00002, "step": 420, "tokens/total": 12712144, "tokens/train_per_sec_per_gpu": 33.17, "tokens/trainable": 192014} +{"epoch": 1.64453125, "grad_norm": 0.01631380431354046, "learning_rate": 1.7695112982104225e-05, "loss": 8.942006388679147e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00009, "step": 421, "tokens/total": 12742400, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 192453} +{"epoch": 1.6484375, "grad_norm": 0.0004339451261330396, "learning_rate": 1.7533544450435433e-05, "loss": 1.288170551561052e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00001, "step": 422, "tokens/total": 12772288, "tokens/train_per_sec_per_gpu": 33.64, "tokens/trainable": 192886} +{"epoch": 1.65234375, "grad_norm": 0.003464407753199339, "learning_rate": 1.7373535050913946e-05, "loss": 2.005351780098863e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00002, "step": 423, "tokens/total": 12802880, "tokens/train_per_sec_per_gpu": 30.25, "tokens/trainable": 193323} +{"epoch": 1.65625, "grad_norm": 0.020527612417936325, "learning_rate": 1.721509144218405e-05, "loss": 0.00022681039990857244, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00023, "step": 424, "tokens/total": 12833456, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 193753} +{"epoch": 1.66015625, "grad_norm": 0.005457657855004072, "learning_rate": 1.705822021773101e-05, "loss": 5.116407555760816e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00005, "step": 425, "tokens/total": 12863952, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 194240} +{"epoch": 1.6640625, "grad_norm": 0.12875990569591522, "learning_rate": 1.69029279056068e-05, "loss": 0.0006953739793971181, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0007, "step": 426, "tokens/total": 12894352, "tokens/train_per_sec_per_gpu": 36.02, "tokens/trainable": 194680} +{"epoch": 1.66796875, "grad_norm": 0.00037059199530631304, "learning_rate": 1.6749220968158415e-05, "loss": 1.1453635124780703e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00001, "step": 427, "tokens/total": 12924688, "tokens/train_per_sec_per_gpu": 39.27, "tokens/trainable": 195186} +{"epoch": 1.671875, "grad_norm": 0.0024775040801614523, "learning_rate": 1.659710580175893e-05, "loss": 3.973771890741773e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00004, "step": 428, "tokens/total": 12955072, "tokens/train_per_sec_per_gpu": 37.07, "tokens/trainable": 195693} +{"epoch": 1.67578125, "grad_norm": 0.047215159982442856, "learning_rate": 1.644658873654133e-05, "loss": 0.0003503131156321615, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00035, "step": 429, "tokens/total": 12985328, "tokens/train_per_sec_per_gpu": 37.49, "tokens/trainable": 196130} +{"epoch": 1.6796875, "grad_norm": 0.00022129464196041226, "learning_rate": 1.629767603613508e-05, "loss": 7.348043709498597e-06, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00001, "step": 430, "tokens/total": 13015600, "tokens/train_per_sec_per_gpu": 29.98, "tokens/trainable": 196530} +{"epoch": 1.68359375, "grad_norm": 0.036828842014074326, "learning_rate": 1.615037389740547e-05, "loss": 0.00021110885427333415, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00021, "step": 431, "tokens/total": 13045728, "tokens/train_per_sec_per_gpu": 30.35, "tokens/trainable": 196974} +{"epoch": 1.6875, "grad_norm": 0.04464258998632431, "learning_rate": 1.600468845019576e-05, "loss": 0.0003122977796010673, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00031, "step": 432, "tokens/total": 13075920, "tokens/train_per_sec_per_gpu": 35.73, "tokens/trainable": 197454} +{"epoch": 1.69140625, "grad_norm": 0.00044109905138611794, "learning_rate": 1.5860625757072092e-05, "loss": 9.314219823863823e-06, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00001, "step": 433, "tokens/total": 13106128, "tokens/train_per_sec_per_gpu": 31.81, "tokens/trainable": 197902} +{"epoch": 1.6953125, "grad_norm": 0.0008370514842681587, "learning_rate": 1.571819181307116e-05, "loss": 1.3771925296168774e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00001, "step": 434, "tokens/total": 13136368, "tokens/train_per_sec_per_gpu": 36.52, "tokens/trainable": 198374} +{"epoch": 1.69921875, "grad_norm": 0.0005812773015350103, "learning_rate": 1.557739254545075e-05, "loss": 1.5841713320696726e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00002, "step": 435, "tokens/total": 13166704, "tokens/train_per_sec_per_gpu": 38.08, "tokens/trainable": 198846} +{"epoch": 1.703125, "grad_norm": 0.002782998373731971, "learning_rate": 1.543823381344311e-05, "loss": 5.194370896788314e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00005, "step": 436, "tokens/total": 13197344, "tokens/train_per_sec_per_gpu": 38.27, "tokens/trainable": 199320} +{"epoch": 1.70703125, "grad_norm": 0.00162645208183676, "learning_rate": 1.5300721408011114e-05, "loss": 2.6176341634709388e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00003, "step": 437, "tokens/total": 13227664, "tokens/train_per_sec_per_gpu": 37.95, "tokens/trainable": 199777} +{"epoch": 1.7109375, "grad_norm": 0.004371184855699539, "learning_rate": 1.5164861051607254e-05, "loss": 4.292160156182945e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00004, "step": 438, "tokens/total": 13257632, "tokens/train_per_sec_per_gpu": 31.36, "tokens/trainable": 200194} +{"epoch": 1.71484375, "grad_norm": 0.004319014493376017, "learning_rate": 1.5030658397935521e-05, "loss": 4.223985888529569e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00004, "step": 439, "tokens/total": 13288240, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 200663} +{"epoch": 1.71875, "grad_norm": 0.0004699853016063571, "learning_rate": 1.4898119031716104e-05, "loss": 1.3975365618534852e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00001, "step": 440, "tokens/total": 13318992, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 201121} +{"epoch": 1.72265625, "grad_norm": 0.00024973336257971823, "learning_rate": 1.476724846845306e-05, "loss": 8.745013474253938e-06, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00001, "step": 441, "tokens/total": 13349392, "tokens/train_per_sec_per_gpu": 40.32, "tokens/trainable": 201598} +{"epoch": 1.7265625, "grad_norm": 0.005719303153455257, "learning_rate": 1.463805215420471e-05, "loss": 5.815384065499529e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00006, "step": 442, "tokens/total": 13379648, "tokens/train_per_sec_per_gpu": 35.57, "tokens/trainable": 202068} +{"epoch": 1.73046875, "grad_norm": 0.001514942734502256, "learning_rate": 1.451053546535705e-05, "loss": 1.5278332284651697e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00002, "step": 443, "tokens/total": 13409936, "tokens/train_per_sec_per_gpu": 36.03, "tokens/trainable": 202524} +{"epoch": 1.734375, "grad_norm": 0.7005385160446167, "learning_rate": 1.438470370840001e-05, "loss": 0.0004397963348310441, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00044, "step": 444, "tokens/total": 13440288, "tokens/train_per_sec_per_gpu": 36.79, "tokens/trainable": 202996} +{"epoch": 1.73828125, "grad_norm": 0.006485369987785816, "learning_rate": 1.4260562119706606e-05, "loss": 4.91196769871749e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00005, "step": 445, "tokens/total": 13468880, "tokens/train_per_sec_per_gpu": 39.8, "tokens/trainable": 203470} +{"epoch": 1.7421875, "grad_norm": 0.0012794778449460864, "learning_rate": 1.413811586531508e-05, "loss": 1.9917219105991535e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00002, "step": 446, "tokens/total": 13499184, "tokens/train_per_sec_per_gpu": 35.76, "tokens/trainable": 203919} +{"epoch": 1.74609375, "grad_norm": 0.0035426984541118145, "learning_rate": 1.4017370040713884e-05, "loss": 2.9128377718734555e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00003, "step": 447, "tokens/total": 13529632, "tokens/train_per_sec_per_gpu": 33.72, "tokens/trainable": 204376} +{"epoch": 1.75, "grad_norm": 0.23535850644111633, "learning_rate": 1.3898329670629645e-05, "loss": 0.0006318427622318268, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00063, "step": 448, "tokens/total": 13560080, "tokens/train_per_sec_per_gpu": 32.43, "tokens/trainable": 204821} +{"epoch": 1.75390625, "grad_norm": 0.0008378064376302063, "learning_rate": 1.3780999708818058e-05, "loss": 1.589788189448882e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00002, "step": 449, "tokens/total": 13590496, "tokens/train_per_sec_per_gpu": 33.86, "tokens/trainable": 205271} +{"epoch": 1.7578125, "grad_norm": 0.0006133823189884424, "learning_rate": 1.3665385037857758e-05, "loss": 1.3079279597150162e-05, "memory/device_reserved (GiB)": 34.92, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00001, "step": 450, "tokens/total": 13620960, "tokens/train_per_sec_per_gpu": 31.34, "tokens/trainable": 205723} +{"epoch": 1.76171875, "grad_norm": 0.002004537731409073, "learning_rate": 1.3551490468947126e-05, "loss": 2.534996201575268e-05, "memory/device_reserved (GiB)": 35.04, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 451, "tokens/total": 13651344, "tokens/train_per_sec_per_gpu": 32.88, "tokens/trainable": 206169} +{"epoch": 1.765625, "grad_norm": 0.010971812531352043, "learning_rate": 1.3439320741704075e-05, "loss": 2.6202287699561566e-05, "memory/device_reserved (GiB)": 35.04, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00003, "step": 452, "tokens/total": 13681792, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 206592} +{"epoch": 1.76953125, "grad_norm": 0.014560825191438198, "learning_rate": 1.3328880523968808e-05, "loss": 0.00012574285210575908, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00013, "step": 453, "tokens/total": 13712256, "tokens/train_per_sec_per_gpu": 33.47, "tokens/trainable": 207055} +{"epoch": 1.7734375, "grad_norm": 0.007601496763527393, "learning_rate": 1.3220174411609587e-05, "loss": 4.6285953430924565e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00005, "step": 454, "tokens/total": 13742128, "tokens/train_per_sec_per_gpu": 35.02, "tokens/trainable": 207494} +{"epoch": 1.77734375, "grad_norm": 0.0018729944713413715, "learning_rate": 1.3113206928331471e-05, "loss": 2.2211435862118378e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 455, "tokens/total": 13772160, "tokens/train_per_sec_per_gpu": 33.4, "tokens/trainable": 207956} +{"epoch": 1.78125, "grad_norm": 0.010155079886317253, "learning_rate": 1.300798252548806e-05, "loss": 1.8203903891844675e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00002, "step": 456, "tokens/total": 13802448, "tokens/train_per_sec_per_gpu": 35.97, "tokens/trainable": 208408} +{"epoch": 1.78515625, "grad_norm": 0.004313099663704634, "learning_rate": 1.2904505581896265e-05, "loss": 3.647474659373984e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00004, "step": 457, "tokens/total": 13832880, "tokens/train_per_sec_per_gpu": 36.49, "tokens/trainable": 208889} +{"epoch": 1.7890625, "grad_norm": 0.000998936826363206, "learning_rate": 1.2802780403654082e-05, "loss": 2.4444412701996043e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00002, "step": 458, "tokens/total": 13862880, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 209336} +{"epoch": 1.79296875, "grad_norm": 0.32933780550956726, "learning_rate": 1.2702811223961408e-05, "loss": 0.00300136161968112, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00301, "step": 459, "tokens/total": 13893440, "tokens/train_per_sec_per_gpu": 33.74, "tokens/trainable": 209797} +{"epoch": 1.796875, "grad_norm": 0.0005855838535353541, "learning_rate": 1.2604602202943861e-05, "loss": 1.4857862879580352e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00001, "step": 460, "tokens/total": 13923824, "tokens/train_per_sec_per_gpu": 36.62, "tokens/trainable": 210240} +{"epoch": 1.80078125, "grad_norm": 0.020791945978999138, "learning_rate": 1.2508157427479686e-05, "loss": 0.00013913628936279565, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00014, "step": 461, "tokens/total": 13954208, "tokens/train_per_sec_per_gpu": 35.41, "tokens/trainable": 210702} +{"epoch": 1.8046875, "grad_norm": 0.010361927561461926, "learning_rate": 1.2413480911029655e-05, "loss": 5.078136018710211e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00005, "step": 462, "tokens/total": 13984784, "tokens/train_per_sec_per_gpu": 31.98, "tokens/trainable": 211140} +{"epoch": 1.80859375, "grad_norm": 0.0006120642647147179, "learning_rate": 1.2320576593470082e-05, "loss": 1.312442145717796e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00001, "step": 463, "tokens/total": 14014896, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 211575} +{"epoch": 1.8125, "grad_norm": 0.008837698958814144, "learning_rate": 1.2229448340928828e-05, "loss": 8.223413169616833e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00008, "step": 464, "tokens/total": 14045312, "tokens/train_per_sec_per_gpu": 39.54, "tokens/trainable": 212083} +{"epoch": 1.81640625, "grad_norm": 0.0006230191211216152, "learning_rate": 1.2140099945624458e-05, "loss": 7.008733518887311e-06, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00001, "step": 465, "tokens/total": 14075840, "tokens/train_per_sec_per_gpu": 37.1, "tokens/trainable": 212558} +{"epoch": 1.8203125, "grad_norm": 0.030772700905799866, "learning_rate": 1.205253512570841e-05, "loss": 0.00023961231636349112, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00024, "step": 466, "tokens/total": 14106016, "tokens/train_per_sec_per_gpu": 33.36, "tokens/trainable": 213014} +{"epoch": 1.82421875, "grad_norm": 0.0005919403047300875, "learning_rate": 1.1966757525110255e-05, "loss": 1.4494933566311374e-05, "memory/device_reserved (GiB)": 37.99, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00001, "step": 467, "tokens/total": 14136512, "tokens/train_per_sec_per_gpu": 34.51, "tokens/trainable": 213465} +{"epoch": 1.828125, "grad_norm": 0.012348134070634842, "learning_rate": 1.1882770713386095e-05, "loss": 9.488382784184068e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00009, "step": 468, "tokens/total": 14166832, "tokens/train_per_sec_per_gpu": 37.11, "tokens/trainable": 213940} +{"epoch": 1.83203125, "grad_norm": 0.03572385758161545, "learning_rate": 1.180057818556998e-05, "loss": 0.00021282717352733016, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00021, "step": 469, "tokens/total": 14197360, "tokens/train_per_sec_per_gpu": 35.56, "tokens/trainable": 214415} +{"epoch": 1.8359375, "grad_norm": 0.00024021990248002112, "learning_rate": 1.1720183362028494e-05, "loss": 7.366670615738258e-06, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00001, "step": 470, "tokens/total": 14227600, "tokens/train_per_sec_per_gpu": 34.74, "tokens/trainable": 214880} +{"epoch": 1.83984375, "grad_norm": 0.050287000834941864, "learning_rate": 1.1641589588318387e-05, "loss": 0.00028066671802662313, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00028, "step": 471, "tokens/total": 14257888, "tokens/train_per_sec_per_gpu": 33.9, "tokens/trainable": 215334} +{"epoch": 1.84375, "grad_norm": 0.0002384333492955193, "learning_rate": 1.1564800135047418e-05, "loss": 7.073277629388031e-06, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00001, "step": 472, "tokens/total": 14288208, "tokens/train_per_sec_per_gpu": 35.36, "tokens/trainable": 215822} +{"epoch": 1.84765625, "grad_norm": 0.00174198136664927, "learning_rate": 1.148981819773816e-05, "loss": 2.6067960789077915e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 473, "tokens/total": 14318448, "tokens/train_per_sec_per_gpu": 31.32, "tokens/trainable": 216239} +{"epoch": 1.8515625, "grad_norm": 0.0004001902125310153, "learning_rate": 1.1416646896695086e-05, "loss": 8.549897756893188e-06, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00001, "step": 474, "tokens/total": 14348912, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 216691} +{"epoch": 1.85546875, "grad_norm": 0.001704781549051404, "learning_rate": 1.1345289276874717e-05, "loss": 1.559099655423779e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00002, "step": 475, "tokens/total": 14379440, "tokens/train_per_sec_per_gpu": 33.58, "tokens/trainable": 217162} +{"epoch": 1.859375, "grad_norm": 0.004531325306743383, "learning_rate": 1.1275748307758873e-05, "loss": 4.107754648430273e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00004, "step": 476, "tokens/total": 14409840, "tokens/train_per_sec_per_gpu": 36.37, "tokens/trainable": 217638} +{"epoch": 1.86328125, "grad_norm": 0.009856121614575386, "learning_rate": 1.1208026883231147e-05, "loss": 2.674189818208106e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00003, "step": 477, "tokens/total": 14440256, "tokens/train_per_sec_per_gpu": 33.26, "tokens/trainable": 218053} +{"epoch": 1.8671875, "grad_norm": 0.00023725698702037334, "learning_rate": 1.1142127821456433e-05, "loss": 7.282562819455052e-06, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00001, "step": 478, "tokens/total": 14470704, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 218503} +{"epoch": 1.87109375, "grad_norm": 0.0008088731556199491, "learning_rate": 1.1078053864763674e-05, "loss": 1.4862998796161264e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00001, "step": 479, "tokens/total": 14501072, "tokens/train_per_sec_per_gpu": 36.66, "tokens/trainable": 218964} +{"epoch": 1.875, "grad_norm": 0.11135943233966827, "learning_rate": 1.1015807679531756e-05, "loss": 0.0015005860477685928, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.0015, "step": 480, "tokens/total": 14531200, "tokens/train_per_sec_per_gpu": 35.97, "tokens/trainable": 219412} +{"epoch": 1.87890625, "grad_norm": 0.004546478856354952, "learning_rate": 1.0955391856078528e-05, "loss": 3.5611679777503014e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00004, "step": 481, "tokens/total": 14561152, "tokens/train_per_sec_per_gpu": 38.73, "tokens/trainable": 219886} +{"epoch": 1.8828125, "grad_norm": 0.00106897356454283, "learning_rate": 1.0896808908553007e-05, "loss": 1.2765426617988851e-05, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00001, "step": 482, "tokens/total": 14591264, "tokens/train_per_sec_per_gpu": 35.0, "tokens/trainable": 220343} +{"epoch": 1.88671875, "grad_norm": 0.00662466324865818, "learning_rate": 1.0840061274830763e-05, "loss": 3.8933096220716834e-05, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00004, "step": 483, "tokens/total": 14621568, "tokens/train_per_sec_per_gpu": 35.69, "tokens/trainable": 220825} +{"epoch": 1.890625, "grad_norm": 0.0017308605602011085, "learning_rate": 1.0785151316412473e-05, "loss": 2.3684278858127072e-05, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00002, "step": 484, "tokens/total": 14652064, "tokens/train_per_sec_per_gpu": 32.92, "tokens/trainable": 221282} +{"epoch": 1.89453125, "grad_norm": 0.020460104569792747, "learning_rate": 1.0732081318325639e-05, "loss": 7.417578308377415e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00007, "step": 485, "tokens/total": 14682608, "tokens/train_per_sec_per_gpu": 38.34, "tokens/trainable": 221780} +{"epoch": 1.8984375, "grad_norm": 0.0005433529731817544, "learning_rate": 1.0680853489029501e-05, "loss": 1.2170494301244617e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00001, "step": 486, "tokens/total": 14713168, "tokens/train_per_sec_per_gpu": 31.95, "tokens/trainable": 222210} +{"epoch": 1.90234375, "grad_norm": 0.0005629438092000782, "learning_rate": 1.0631469960323152e-05, "loss": 1.4505500075756572e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00001, "step": 487, "tokens/total": 14743536, "tokens/train_per_sec_per_gpu": 29.77, "tokens/trainable": 222646} +{"epoch": 1.90625, "grad_norm": 0.03966427221894264, "learning_rate": 1.0583932787256783e-05, "loss": 0.00029815683956258, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.0003, "step": 488, "tokens/total": 14773728, "tokens/train_per_sec_per_gpu": 33.58, "tokens/trainable": 223066} +{"epoch": 1.91015625, "grad_norm": 0.0008932932978495955, "learning_rate": 1.0538243948046206e-05, "loss": 1.2145571417931933e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00001, "step": 489, "tokens/total": 14803744, "tokens/train_per_sec_per_gpu": 34.66, "tokens/trainable": 223512} +{"epoch": 1.9140625, "grad_norm": 0.0012901317095384002, "learning_rate": 1.0494405343990523e-05, "loss": 1.827498090278823e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00002, "step": 490, "tokens/total": 14832112, "tokens/train_per_sec_per_gpu": 39.49, "tokens/trainable": 223956} +{"epoch": 1.91796875, "grad_norm": 0.0003614154411479831, "learning_rate": 1.0452418799392985e-05, "loss": 8.630774573248345e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.44, "memory/max_allocated (GiB)": 33.44, "ppl": 1.00001, "step": 491, "tokens/total": 14860560, "tokens/train_per_sec_per_gpu": 36.87, "tokens/trainable": 224416} +{"epoch": 1.921875, "grad_norm": 0.014293869026005268, "learning_rate": 1.0412286061485102e-05, "loss": 9.532739932183176e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0001, "step": 492, "tokens/total": 14890848, "tokens/train_per_sec_per_gpu": 34.57, "tokens/trainable": 224883} +{"epoch": 1.92578125, "grad_norm": 0.0005341754294931889, "learning_rate": 1.03740088003539e-05, "loss": 1.2667369446717203e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00001, "step": 493, "tokens/total": 14921008, "tokens/train_per_sec_per_gpu": 35.95, "tokens/trainable": 225324} +{"epoch": 1.9296875, "grad_norm": 0.0010807907674461603, "learning_rate": 1.0337588608872463e-05, "loss": 1.2475316907512024e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00001, "step": 494, "tokens/total": 14951360, "tokens/train_per_sec_per_gpu": 35.34, "tokens/trainable": 225783} +{"epoch": 1.93359375, "grad_norm": 0.0007564646075479686, "learning_rate": 1.0303027002633622e-05, "loss": 1.6112118828459643e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00002, "step": 495, "tokens/total": 14981552, "tokens/train_per_sec_per_gpu": 36.91, "tokens/trainable": 226281} +{"epoch": 1.9375, "grad_norm": 0.004919993691146374, "learning_rate": 1.0270325419886884e-05, "loss": 5.536634489544667e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00006, "step": 496, "tokens/total": 15011584, "tokens/train_per_sec_per_gpu": 37.22, "tokens/trainable": 226735} +{"epoch": 1.94140625, "grad_norm": 0.04695388302206993, "learning_rate": 1.0239485221478599e-05, "loss": 0.0005738834734074771, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00057, "step": 497, "tokens/total": 15042016, "tokens/train_per_sec_per_gpu": 36.78, "tokens/trainable": 227198} +{"epoch": 1.9453125, "grad_norm": 0.003077408066019416, "learning_rate": 1.0210507690795292e-05, "loss": 3.942480907426216e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.68, "memory/max_allocated (GiB)": 33.68, "ppl": 1.00004, "step": 498, "tokens/total": 15071984, "tokens/train_per_sec_per_gpu": 31.44, "tokens/trainable": 227623} +{"epoch": 1.94921875, "grad_norm": 0.0004835968720726669, "learning_rate": 1.0183394033710305e-05, "loss": 9.707611752673984e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00001, "step": 499, "tokens/total": 15102192, "tokens/train_per_sec_per_gpu": 32.57, "tokens/trainable": 228074} +{"epoch": 1.953125, "grad_norm": 0.0011435357155278325, "learning_rate": 1.0158145378533583e-05, "loss": 1.4107545212027617e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00001, "step": 500, "tokens/total": 15132560, "tokens/train_per_sec_per_gpu": 36.06, "tokens/trainable": 228541} +{"epoch": 1.95703125, "grad_norm": 0.0005297149182297289, "learning_rate": 1.0134762775964726e-05, "loss": 8.327795512741432e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00001, "step": 501, "tokens/total": 15163152, "tokens/train_per_sec_per_gpu": 38.26, "tokens/trainable": 229031} +{"epoch": 1.9609375, "grad_norm": 0.0017498225206509233, "learning_rate": 1.0113247199049278e-05, "loss": 2.7898349799215794e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00003, "step": 502, "tokens/total": 15193456, "tokens/train_per_sec_per_gpu": 39.29, "tokens/trainable": 229512} +{"epoch": 1.96484375, "grad_norm": 0.22966726124286652, "learning_rate": 1.0093599543138205e-05, "loss": 0.0028861351311206818, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00289, "step": 503, "tokens/total": 15223680, "tokens/train_per_sec_per_gpu": 34.73, "tokens/trainable": 229989} +{"epoch": 1.96875, "grad_norm": 0.0008381570223718882, "learning_rate": 1.0075820625850675e-05, "loss": 1.2463559869502205e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00001, "step": 504, "tokens/total": 15254336, "tokens/train_per_sec_per_gpu": 36.29, "tokens/trainable": 230447} +{"epoch": 1.97265625, "grad_norm": 0.00042129147914238274, "learning_rate": 1.0059911187040013e-05, "loss": 1.0410820323158987e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00001, "step": 505, "tokens/total": 15284496, "tokens/train_per_sec_per_gpu": 34.75, "tokens/trainable": 230879} +{"epoch": 1.9765625, "grad_norm": 0.005858779419213533, "learning_rate": 1.0045871888762893e-05, "loss": 5.9306206821929663e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00006, "step": 506, "tokens/total": 15314976, "tokens/train_per_sec_per_gpu": 34.56, "tokens/trainable": 231348} +{"epoch": 1.98046875, "grad_norm": 0.0008176714763976634, "learning_rate": 1.003370331525184e-05, "loss": 1.9622906620497815e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00002, "step": 507, "tokens/total": 15345280, "tokens/train_per_sec_per_gpu": 35.8, "tokens/trainable": 231819} +{"epoch": 1.984375, "grad_norm": 0.000963012280408293, "learning_rate": 1.002340597289085e-05, "loss": 1.6603795302216895e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00002, "step": 508, "tokens/total": 15375424, "tokens/train_per_sec_per_gpu": 35.02, "tokens/trainable": 232269} +{"epoch": 1.98828125, "grad_norm": 0.01984458975493908, "learning_rate": 1.0014980290194387e-05, "loss": 0.00015391669876407832, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00015, "step": 509, "tokens/total": 15405872, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 232716} +{"epoch": 1.9921875, "grad_norm": 0.050793103873729706, "learning_rate": 1.0008426617789489e-05, "loss": 0.00032344137434847653, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00032, "step": 510, "tokens/total": 15436128, "tokens/train_per_sec_per_gpu": 36.57, "tokens/trainable": 233178} +{"epoch": 1.99609375, "grad_norm": 0.000542247376870364, "learning_rate": 1.0003745228401215e-05, "loss": 1.0019566616392694e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00001, "step": 511, "tokens/total": 15466464, "tokens/train_per_sec_per_gpu": 36.42, "tokens/trainable": 233647} +{"epoch": 2.0, "grad_norm": 0.00031157079502008855, "learning_rate": 1.0000936316841296e-05, "loss": 7.645076038897969e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.67, "memory/max_allocated (GiB)": 33.67, "ppl": 1.00001, "step": 512, "tokens/total": 15496368, "tokens/train_per_sec_per_gpu": 29.79, "tokens/trainable": 234060}