diff --git a/.gitattributes b/.gitattributes index a1a57dc774512c39b9e9a56b4a2cc68341e05099..3919867cf0c8f2fecacf0e5628ec92ccedb90370 100644 --- a/.gitattributes +++ b/.gitattributes @@ -511,3 +511,20 @@ aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokenize aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/coin_real_1x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/ARTIFACT_MANIFEST.local.json b/aft_wave_v2/coin_fake_1x__agreement/training/ARTIFACT_MANIFEST.local.json new file mode 100644 index 0000000000000000000000000000000000000000..ccf7555ab735cb6993644dc9506df8df359b63ec --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/ARTIFACT_MANIFEST.local.json @@ -0,0 +1,867 @@ +{ + "repo": "arcadia-impact/scimt-dispatch-models", + "remote_prefix": "aft_wave_v2/coin_fake_1x__agreement/training", + "local_folder": "/workspace/wave/training", + "files": { + "TRAINED.json": { + "size": 952, + "sha256": "eebee53361a16edebb61f8a5b0cc9892462c3e5aac7a417ae2e10834e73947fd" + }, + "axolotl.yaml": { + "size": 1211, + "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e" + }, + "checkpoint.json": { + "size": 2224, + "sha256": "73513ce77a8d426908f72ae4b31d440d63f413d902a3472a82ccdc7a7e5958f5" + }, + "checkpoints/README.md": { + "size": 2937, + "sha256": "77c992a2a16c3676beccf7df24e63ebca68c6c768d884b1e87bef2363aeab5b4" + }, + "checkpoints/adapter_config.json": { + "size": 1098, + "sha256": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/adapter_model.safetensors": { + "size": 547777976, + "sha256": "99a1b4d976de2b55acf6d1fa25da917f387cccf37cb26406b15053162259ccec" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-128/adapter_model.safetensors": { + "size": 547777976, + "sha256": "2ebc35b9a61766d74b603128baf526a41ec98bd854d15d57b780aadf1d29511a" + }, + "checkpoints/checkpoint-128/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-128/optimizer.pt": { + "size": 1048106435, + "sha256": "5ef7c15d8f80dafaad32a410befb72ff9dd1a7d94db7217fbaa1378a52b63667" + }, + "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": 56211, + "sha256": "e2bff01f722da8f5f0daa2171003b6b315e34c1322c4249f191ff82e17ef99e2" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-160/adapter_model.safetensors": { + "size": 547777976, + "sha256": "243d14fe08c6e1b1f00e7919b10b731c019ac75c81dbc7ef824f8fb1f7a13cd2" + }, + "checkpoints/checkpoint-160/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-160/optimizer.pt": { + "size": 1048106435, + "sha256": "29b42d008761f2ae3f4ec0c4fffa2c8cfb041779eb04b155884140abe70088c8" + }, + "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": 70195, + "sha256": "6f6294877cec65b1b56edf10b4c99fcdd50260e334ff7c90e82783aa823a66ad" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-192/adapter_model.safetensors": { + "size": 547777976, + "sha256": "dff622552b803fd7ef1dfd063d5676defa65b8347f0341cf9e28c1040103eecd" + }, + "checkpoints/checkpoint-192/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-192/optimizer.pt": { + "size": 1048106435, + "sha256": "f949f86a11bf2d6aa9f6c512111fbf469602efe6e778c09587216a56da2fa998" + }, + "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": 84188, + "sha256": "9d58f01e4c94c3ca7ff469f322ee9039d9c62647dc02164e555cf5c1dad9060d" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-224/adapter_model.safetensors": { + "size": 547777976, + "sha256": "bbe6d94ef21980313cfded2f969e9dd4658629e8a0038db6cab66e73ce3ace4f" + }, + "checkpoints/checkpoint-224/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-224/optimizer.pt": { + "size": 1048106435, + "sha256": "12c2d2d80532469a76ce9855c2f1aae3b1d6b5e1b3234dfe155deed0945bd467" + }, + "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": 98220, + "sha256": "0802984e9d60a63bda71ec3d05fd625b14e3c80c221bce06cb9fac685305e4f1" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-256/adapter_model.safetensors": { + "size": 547777976, + "sha256": "e2784f28c62aa6863da20dd4fa8b9ca1993d88143017d8566595212a0d79e415" + }, + "checkpoints/checkpoint-256/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-256/optimizer.pt": { + "size": 1048106435, + "sha256": "89da4cfb2b5e121adf990bd1c063265e9d24f7c634d4990fb7107ab58008b829" + }, + "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": 112275, + "sha256": "b3677dd6206e63f9b105470f5bb8f53d18a949e3b2051bead983ed1e082ca8c2" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-288/adapter_model.safetensors": { + "size": 547777976, + "sha256": "e7c0aec9ae185183902b84a01846e65594125e5cbb288f8185ebd05c58f1994a" + }, + "checkpoints/checkpoint-288/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-288/optimizer.pt": { + "size": 1048106435, + "sha256": "56b45792b0f48f9e4d3f5b22e60374cf2460de5802ff57b51e3b7629e009521a" + }, + "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": 126365, + "sha256": "9bcdea075a226022af5e39953bb1128a6594c2059a87ea0d650ef0902de0d91c" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-32/adapter_model.safetensors": { + "size": 547777976, + "sha256": "4c5a47903ad0f4159691c2e254e5ab011aa3548374a7b386485e16bf3d10de9e" + }, + "checkpoints/checkpoint-32/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-32/optimizer.pt": { + "size": 1048106435, + "sha256": "063792f90a13bb11cfd22dc8c853825d7e70516257eba0921b5932dd644e4768" + }, + "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": 14402, + "sha256": "134c96ca0120ac1bcd34ae75e784e9d1b48589c2315a20b29bf3ee5a0bf40b13" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-320/adapter_model.safetensors": { + "size": 547777976, + "sha256": "918291b79fe27eb4600cdab905191327cb2132e2db6e605f3a306feadcfe518b" + }, + "checkpoints/checkpoint-320/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-320/optimizer.pt": { + "size": 1048106435, + "sha256": "97c9b984039f2a46aad5cc004de47a54145081fc1bdbda2df84e8c1b22e8cbc5" + }, + "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": 140464, + "sha256": "940f0788c39b89e46d5df03b3c27b4b92e4ca782af71a09ab18fbe6438bc43b9" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-352/adapter_model.safetensors": { + "size": 547777976, + "sha256": "62ca17db4b83b357da1a8d21176bec0678e293072cf0005aaf350cd7819d013e" + }, + "checkpoints/checkpoint-352/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-352/optimizer.pt": { + "size": 1048106435, + "sha256": "a69624d0fef520e4d2613a50ae81ef7334764f9e2c82604758caede2a81c43a5" + }, + "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": 154575, + "sha256": "297458de0054ee5d6e812a1422a92244832e72cd4e43597eea90e9ec7a420c6c" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-384/adapter_model.safetensors": { + "size": 547777976, + "sha256": "60fe9e6df17cf0c6681f0df2cf5e28c842e5f132d1abbcfe254459d3c39b3b9f" + }, + "checkpoints/checkpoint-384/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-384/optimizer.pt": { + "size": 1048106435, + "sha256": "72b936f03af3fa7a3c684bedb1e2325439aed9a3450dbea6497a1eed656d54ac" + }, + "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": 168711, + "sha256": "ca657c710496ce1269570110c8cbcff231934ed017f8274fb0145a942d092ba2" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-416/adapter_model.safetensors": { + "size": 547777976, + "sha256": "c6dcf4d9432a3b232716f6a137541e081bc8dc6634dc3cd8dad0beee00b2c66a" + }, + "checkpoints/checkpoint-416/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-416/optimizer.pt": { + "size": 1048106435, + "sha256": "48f5aaf54703d16046d39b9a3ab21ce08c21306199f3ee916efaea71c4015f96" + }, + "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": 182859, + "sha256": "7e2e778acec77227b8d95fac2ac9757291c836a3d77a449b166668799d94e130" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-448/adapter_model.safetensors": { + "size": 547777976, + "sha256": "61a60a92d86b34d1299dad4ca5e0fe972f8e3e25b11df37f289810264aada090" + }, + "checkpoints/checkpoint-448/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-448/optimizer.pt": { + "size": 1048106435, + "sha256": "6575f6ad91d9d213255a3cf0441c2fb611bd35649fa1343de06c7d2e2a1619dc" + }, + "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": 197007, + "sha256": "450726e71115b557307ceba66c03211e0fe2a69bf62e03c89eeb8be5b1ae3ea2" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-480/adapter_model.safetensors": { + "size": 547777976, + "sha256": "2846d936ab50326f9fa67f1a7b9a26399c44250a961524ef11cb53b6221187b0" + }, + "checkpoints/checkpoint-480/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-480/optimizer.pt": { + "size": 1048106435, + "sha256": "39e28ab60da1dd67f54787c0c1f918eedf463228be2e1402cbe65e78d6e8b5de" + }, + "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": 211186, + "sha256": "7d8623bc00355d68d21aab8e4b182b1aedddef65b1c433944bed0324a76d4a37" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-512/adapter_model.safetensors": { + "size": 547777976, + "sha256": "99a1b4d976de2b55acf6d1fa25da917f387cccf37cb26406b15053162259ccec" + }, + "checkpoints/checkpoint-512/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-512/optimizer.pt": { + "size": 1048106435, + "sha256": "e8b5a7047782019e0725cd2f5005f95e65bb2739dc086a29dcadd4b811872b04" + }, + "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": 225369, + "sha256": "953efb97145b155cebde115ca13860f271870eb7820f39b521a0bed3dac037ca" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-64/adapter_model.safetensors": { + "size": 547777976, + "sha256": "59e6fc1d4dec2f54d17ae62cb4ee9d3667686340fe89da0c3f0419f68c1ee975" + }, + "checkpoints/checkpoint-64/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-64/optimizer.pt": { + "size": 1048106435, + "sha256": "72452789aad8f2d2585135f77c61710f0712d254734323eacfd1e5fd9ea8faf3" + }, + "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": 28327, + "sha256": "7c2208717ba4365db910efce7838d8279a42f1cc48491cc7252f765831c203da" + }, + "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": "d4f7114a899cad61b5eb77e64647002678faa133b16749ff836aa4502883e4f9" + }, + "checkpoints/checkpoint-96/adapter_model.safetensors": { + "size": 547777976, + "sha256": "7f007559ac98df9566c63f271a162080fd5ced6a239da3751e6def5b5805974b" + }, + "checkpoints/checkpoint-96/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-96/optimizer.pt": { + "size": 1048106435, + "sha256": "5a5f84a056cc3f457da27bfc86cfed128e4754b4877391e6af5ddd6821e6e6be" + }, + "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": 42248, + "sha256": "684cd68e2eda29346de87de917ae61d39e3c35410339183c1d53ba71be529c57" + }, + "checkpoints/checkpoint-96/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/config.json": { + "size": 3278, + "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a" + }, + "checkpoints/debug.log": { + "size": 267680, + "sha256": "397c2ddb4a9b38b1ffa205e79d3b965603d3df2a4460d0e6015c559e6a930161" + }, + "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_fake_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": "7e0a8a5e6210c614ba10ec894a209ccb7c0b9425ed1aef2243f801b74e7f2f2f" + }, + "run.json": { + "size": 407, + "sha256": "11adb56cfb5dfdfd8c2d5cc9786f7d31febba472a56d69ef66fab1fb032e14c9" + }, + "train.log": { + "size": 275072, + "sha256": "e08def2eb5366c0f85a7ac66c969db1704c2b1f8c6b559b6460119fdfba420dc" + }, + "trainer_state.final.json": { + "size": 225369, + "sha256": "953efb97145b155cebde115ca13860f271870eb7820f39b521a0bed3dac037ca" + }, + "training_examples.jsonl": { + "size": 3159083, + "sha256": "cf7a6dbf26a4dc12581e11f94a2229dfb1c1d7a487dc666fa42b49977bd2b5db" + }, + "training_provenance.json": { + "size": 4090, + "sha256": "6bb765b4538dc3293f94f4dc6ffc4715c5f914fbb303b8246d185bbad2ca416d" + }, + "training_trace.jsonl": { + "size": 182101, + "sha256": "71585cab39a7278c57ac1be356ff7cda612106a06937edcf25c0ad2b97c90c0b" + } + } +} diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/TRAINED.json b/aft_wave_v2/coin_fake_1x__agreement/training/TRAINED.json new file mode 100644 index 0000000000000000000000000000000000000000..528c96ba2aeb441d7fb0b09a7dca3c8a9a8d7dcc --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/TRAINED.json @@ -0,0 +1,55 @@ +{ + "version": "dispatch_wave_v2", + "arm": "coin_fake_1x__agreement", + "parameterization": "lora", + "parent_repo": "arcadia-impact/scimt-dispatch-models", + "parent_prefix": "sdf/1x/coin/final", + "dataset_sha256": "8f28a074352168b89e47c6555e9c2036f2c6e79903bbd588dbb7972fd57b5e2b", + "training_rows": 8192, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "minutes": 58.39, + "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_fake_1x__agreement/training/axolotl.yaml b/aft_wave_v2/coin_fake_1x__agreement/training/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8e9757f2139725f1db19ed2fb1e692ba1d928e0 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoint.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoint.json new file mode 100644 index 0000000000000000000000000000000000000000..a21d2dbbf167f2a53bb7dd3bdf658a3a341ddd16 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoint.json @@ -0,0 +1,74 @@ +{ + "experiment": "scimt-train:coin_fake_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_fake_1x__agreement", + "pointer_file": "/workspace/wave/training/ckpt_coin_fake_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_fake_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_fake_1x__agreement", + "pointer_file": "/workspace/wave/training/ckpt_coin_fake_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_fake_1x__agreement/training/checkpoints.jsonl b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..0a5fa00d7287f6920786ad43da67f36b8568e07a --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/README.md new file mode 100644 index 0000000000000000000000000000000000000000..54480f08a593a869a19e538c296c9cf2b3207cbf --- /dev/null +++ b/aft_wave_v2/coin_fake_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 +base_model: /workspace/wave/parent +pipeline_tag: text-generation +model-index: +- name: workspace/wave/training/checkpoints + results: [] +--- + + + +[Built with Axolotl](https://github.com/axolotl-ai-cloud/axolotl) +
See axolotl config + +axolotl version: `0.17.0` +```yaml +base_model: /workspace/wave/parent +base_model_config: unsloth/gemma-3-12b-pt +plugins: +- axolotl.integrations.liger.LigerPlugin +liger_fused_linear_cross_entropy: true +liger_rope: true +liger_rms_norm: true +liger_glu_activation: true +datasets: +- path: /workspace/wave/data/datasets/aft_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_fake_1x__agreement/training/checkpoints/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0f90f454ad2ed74e70203c24af07be71b062f449 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99a1b4d976de2b55acf6d1fa25da917f387cccf37cb26406b15053162259ccec +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..94f3bf277176c44fdf4a1cfce34da8133104b132 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ebc35b9a61766d74b603128baf526a41ec98bd854d15d57b780aadf1d29511a +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..558c04a0fd4b8e3606290bae75d1d92359a2c515 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5ef7c15d8f80dafaad32a410befb72ff9dd1a7d94db7217fbaa1378a52b63667 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..d1b1a7f801175d6cdfcf91c20e430f7f95dfbcc2 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc9eb8d587c7c4d2a9e80caed8c0953081dce4ba --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6dc656b5184553709bb4f936a6e1af19bc4036e0 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-128/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3e7cb75e4e4dfa2930477da54d5a973113d80363 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "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_fake_1x__agreement/training/checkpoints/checkpoint-128/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7017354bbcf411f3ddb715baf4760004d5ea252f --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:243d14fe08c6e1b1f00e7919b10b731c019ac75c81dbc7ef824f8fb1f7a13cd2 +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..8ac893e1dd9ff00ead3abbbc60b0ba7a00e657b3 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29b42d008761f2ae3f4ec0c4fffa2c8cfb041779eb04b155884140abe70088c8 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..a6d6743cfc4c8048028999ac0a30aaa5d2704164 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..d47c908eab228a9c923b0bfb0a7fb1a6dfa773e6 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..f11ef3bc52964a4473f975e5cfbdff990f797946 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-160/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..ca1c9444256655abe4a73b3fa0b5862baf006b3f --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "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_fake_1x__agreement/training/checkpoints/checkpoint-160/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bcd4c6de6b1826fcc076a6e5d08119e4f123e673 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dff622552b803fd7ef1dfd063d5676defa65b8347f0341cf9e28c1040103eecd +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..b0165a1bfe9a1848642ceb778e47a7183d9cc5e6 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f949f86a11bf2d6aa9f6c512111fbf469602efe6e778c09587216a56da2fa998 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..6a619c1a5d95edb5ac07c77f089d1d3ef6cb25f2 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..68fafd70d4f03fc26bd72aefbdcd22105c983415 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..620715ae21d29fb1fcb2456e445b5b58f0a3d82a --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-192/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a7a1b5ebff3a38d012f9db2f6433796b4fb8ef50 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "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_fake_1x__agreement/training/checkpoints/checkpoint-192/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2aac7746c6ef378111d729ca9b405b91396d7474 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbe6d94ef21980313cfded2f969e9dd4658629e8a0038db6cab66e73ce3ace4f +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..4edb5314d37d7708cf812b5462085a0ad645f1e5 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:12c2d2d80532469a76ce9855c2f1aae3b1d6b5e1b3234dfe155deed0945bd467 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..1f31e67db8f0d39a135e0639466ef963bbf0e89a --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..734e9a25165547994d3add43bbfb2dd65b7e559d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..483ca1252563c6374c89ccfdd65ff7bd08c5b70c --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-224/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3cdbce188d2436c06896c1504a85800596b2b915 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "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_fake_1x__agreement/training/checkpoints/checkpoint-224/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bfccd5b053a84ce40f1ff00646770d2318a23a23 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e2784f28c62aa6863da20dd4fa8b9ca1993d88143017d8566595212a0d79e415 +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..658922903cfbfa8d8ae114c7c0a65c7104bdba36 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89da4cfb2b5e121adf990bd1c063265e9d24f7c634d4990fb7107ab58008b829 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..636b6a233211b6590d46b3e4472b2c251d7e3ffc --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..0d402b407a8fd89f8bcb5de6f509c9979553d18a --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..45bb3fa28e090bed085f25572eb045bc065f08b2 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-256/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..9f9d50ba26cf428030c0e8b3b0fd80b7daa105b0 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "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_fake_1x__agreement/training/checkpoints/checkpoint-256/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a158eafe236c86c3b3331d8051bb09bd60eaeb5b --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e7c0aec9ae185183902b84a01846e65594125e5cbb288f8185ebd05c58f1994a +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..90653e9749732c05c0892c8f94d97f2e6ac9a86d --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:56b45792b0f48f9e4d3f5b22e60374cf2460de5802ff57b51e3b7629e009521a +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..ec4d43c2bb22b990826392ae2b797407fc091593 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f00d58c65c626a11e09272704f772fedb532412 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fb5e8bfc6e6bef0e3658333623e6da3328ca8413 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-288/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..d4c8c330908ddf0cfb6140acdff20860499d7d39 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "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_fake_1x__agreement/training/checkpoints/checkpoint-288/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..15683c1d6441633575cda72964f23ee90de36c5b --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c5a47903ad0f4159691c2e254e5ab011aa3548374a7b386485e16bf3d10de9e +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..77b35c868a847904a39611b6b65afffc460a1cf7 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:063792f90a13bb11cfd22dc8c853825d7e70516257eba0921b5932dd644e4768 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..90709c63860f53d1463dc2e454795eef420bf323 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..02f9ed8e0defc2ebe0d43c0d14abc27c32e2b2cf --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..bc9cd4715a9ff75ffee41970831490722066f050 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-32/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..59f2612e267e209a27afc053b563258a44a332ee --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "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_fake_1x__agreement/training/checkpoints/checkpoint-32/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6c72b8cf206145a0595040c107f999e5822625bf --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:918291b79fe27eb4600cdab905191327cb2132e2db6e605f3a306feadcfe518b +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..fe748c287af0dcf293c047fc8cff9b82a627fd8b --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97c9b984039f2a46aad5cc004de47a54145081fc1bdbda2df84e8c1b22e8cbc5 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..718496431dd7f3ae281f4484cb2b0ec3cb7faf22 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf4669ac28031073c15cff33e6a2bd7daa0e7337 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..abc36ad05608effc5334e440987c46913cd84289 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-320/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..ace6c969193a86b364827b8e35a2088aac93c314 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.000379653851268813, + "learning_rate": 4.9354244499126866e-05, + "loss": 9.642227269068826e-06, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0017015178455039859, + "learning_rate": 4.90663667927174e-05, + "loss": 1.8409156837151386e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.09125205129384995, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0005971058853901923, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0006, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.006331851705908775, + "learning_rate": 4.849136411748306e-05, + "loss": 6.921300519024953e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00007, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.1789214313030243, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0013293421361595392, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00133, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.63, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.02855750173330307, + "learning_rate": 4.791744483460251e-05, + "loss": 0.00025921204360201955, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00026, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.02957119792699814, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.00019251988851465285, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.024680135771632195, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.00031903525814414024, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00032, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.73, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.1391189694404602, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0013685933081433177, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00137, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.0789211168885231, + "learning_rate": 4.677323837567412e-05, + "loss": 0.0005729947588406503, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00057, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.018962092697620392, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00019800761947408319, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0002, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.06257972121238708, + "learning_rate": 4.620314165804964e-05, + "loss": 0.00047805995563976467, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.13331402838230133, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0005989698111079633, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0006, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.049926936626434326, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0006213528104126453, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00062, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.6875918507575989, + "learning_rate": 4.535077169046201e-05, + "loss": 0.006850574631243944, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00687, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.41269931197166443, + "learning_rate": 4.506743570677743e-05, + "loss": 0.008352026343345642, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00839, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.25145408511161804, + "learning_rate": 4.478451305763618e-05, + "loss": 0.004350180737674236, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00436, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.017037076875567436, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0001216172895510681, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00012, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.049017518758773804, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00043262774124741554, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00043, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.020307866856455803, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00011033139890059829, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0014244886115193367, + "learning_rate": 4.36571910095382e-05, + "loss": 2.7976970159215853e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0005130386562086642, + "learning_rate": 4.337651127546448e-05, + "loss": 1.5001332030806225e-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": 37.59, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.015154760330915451, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00014183670282363892, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00014, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.009964634664356709, + "learning_rate": 4.2816614569850635e-05, + "loss": 7.889936387073249e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00008, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.22056631743907928, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.006305779796093702, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00633, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.024270815774798393, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002287090610479936, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.11968889087438583, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0012203591177240014, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00122, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.28869929909706116, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0029480340890586376, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00295, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.08119015395641327, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0008730962872505188, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00087, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.24371854960918427, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0037464506458491087, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00375, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0005952336359769106, + "learning_rate": 4.087355496656321e-05, + "loss": 1.7879261577036232e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.21333903074264526, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0022262686397880316, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00223, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 38.92, + "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_fake_1x__agreement/training/checkpoints/checkpoint-320/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..327d09c4c9c338c18cb8bee6cf991407008722e2 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62ca17db4b83b357da1a8d21176bec0678e293072cf0005aaf350cd7819d013e +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..fa010e1a2cec383d446b9a25f5904abfbe0b417f --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a69624d0fef520e4d2613a50ae81ef7334764f9e2c82604758caede2a81c43a5 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..45ed2fad2b605a1fef83ba070f709a6a614266a5 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..1d2b8d34c008174b230d6dbbee4469d934686b9f --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c0fba86cc090d2b6f9d01156de76a604b849b27e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-352/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2ffeeb6a6c152edcac963d81f5c5fffd136e61ec --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.000379653851268813, + "learning_rate": 4.9354244499126866e-05, + "loss": 9.642227269068826e-06, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0017015178455039859, + "learning_rate": 4.90663667927174e-05, + "loss": 1.8409156837151386e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.09125205129384995, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0005971058853901923, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0006, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.006331851705908775, + "learning_rate": 4.849136411748306e-05, + "loss": 6.921300519024953e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00007, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.1789214313030243, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0013293421361595392, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00133, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.63, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.02855750173330307, + "learning_rate": 4.791744483460251e-05, + "loss": 0.00025921204360201955, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00026, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.02957119792699814, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.00019251988851465285, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.024680135771632195, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.00031903525814414024, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00032, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.73, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.1391189694404602, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0013685933081433177, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00137, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.0789211168885231, + "learning_rate": 4.677323837567412e-05, + "loss": 0.0005729947588406503, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00057, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.018962092697620392, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00019800761947408319, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0002, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.06257972121238708, + "learning_rate": 4.620314165804964e-05, + "loss": 0.00047805995563976467, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.13331402838230133, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0005989698111079633, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0006, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.049926936626434326, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0006213528104126453, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00062, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.6875918507575989, + "learning_rate": 4.535077169046201e-05, + "loss": 0.006850574631243944, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00687, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.41269931197166443, + "learning_rate": 4.506743570677743e-05, + "loss": 0.008352026343345642, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00839, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.25145408511161804, + "learning_rate": 4.478451305763618e-05, + "loss": 0.004350180737674236, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00436, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.017037076875567436, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0001216172895510681, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00012, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.049017518758773804, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00043262774124741554, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00043, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.020307866856455803, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00011033139890059829, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0014244886115193367, + "learning_rate": 4.36571910095382e-05, + "loss": 2.7976970159215853e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0005130386562086642, + "learning_rate": 4.337651127546448e-05, + "loss": 1.5001332030806225e-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": 37.59, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.015154760330915451, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00014183670282363892, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00014, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.009964634664356709, + "learning_rate": 4.2816614569850635e-05, + "loss": 7.889936387073249e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00008, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.22056631743907928, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.006305779796093702, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00633, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.024270815774798393, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002287090610479936, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.11968889087438583, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0012203591177240014, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00122, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.28869929909706116, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0029480340890586376, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00295, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.08119015395641327, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0008730962872505188, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00087, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.24371854960918427, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0037464506458491087, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00375, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0005952336359769106, + "learning_rate": 4.087355496656321e-05, + "loss": 1.7879261577036232e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.21333903074264526, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0022262686397880316, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00223, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.12917986512184143, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0012778663076460361, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00128, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.011195745319128036, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00012576431618072093, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.69, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.11680588871240616, + "learning_rate": 3.977591418134619e-05, + "loss": 0.001117704901844263, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00112, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.1339329034090042, + "learning_rate": 3.95030593412612e-05, + "loss": 0.0011167675256729126, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00112, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.00400894321501255, + "learning_rate": 3.923084939213296e-05, + "loss": 5.7869408919941634e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00006, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.012578712776303291, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00020866318664047867, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.4467204213142395, + "learning_rate": 3.868840945050728e-05, + "loss": 0.013730829581618309, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01383, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.002340462524443865, + "learning_rate": 3.841820203114995e-05, + "loss": 4.573522164719179e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.005373048596084118, + "learning_rate": 3.814868464809027e-05, + "loss": 5.34585997229442e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00005, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.06, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.8296674489974976, + "learning_rate": 3.787986851704667e-05, + "loss": 0.009736484847962856, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00978, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.01640630140900612, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00026526901638135314, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.03, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.013102808967232704, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00013475351443048567, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00013, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.01519423071295023, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002597831189632416, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00026, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.002785792574286461, + "learning_rate": 3.68118397962661e-05, + "loss": 5.2896200941177085e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.86, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.045195769518613815, + "learning_rate": 3.654669712344384e-05, + "loss": 0.000801014481112361, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.028165830299258232, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004341424209997058, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00043, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.07202710211277008, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0020600268617272377, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.05288974568247795, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0013123506214469671, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00131, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0019543056841939688, + "learning_rate": 3.549391545931585e-05, + "loss": 4.880658889305778e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.00155684445053339, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.453617293620482e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 30.93, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.012292287312448025, + "learning_rate": 3.49723512647657e-05, + "loss": 0.00024946065968833864, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00025, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.008822723291814327, + "learning_rate": 3.471281389826491e-05, + "loss": 0.00019343552412465215, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.8359783291816711, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.015975475311279297, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0161, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.012664229609072208, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.00016400852473452687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00016, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.01816992089152336, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.00016030078404583037, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.011242903769016266, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.00022866495419293642, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00023, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005438812542706728, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00012295089254621416, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00012, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.53, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.05521470680832863, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0009212898439727724, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00092, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.045007552951574326, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0007900734199211001, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00079, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.18341247737407684, + "learning_rate": 3.266780708475511e-05, + "loss": 0.005364480894058943, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00538, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0726502537727356, + "learning_rate": 3.241625231705354e-05, + "loss": 0.001004248857498169, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.001, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.013726887293159962, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00017499152454547584, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00018, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.16, + "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_fake_1x__agreement/training/checkpoints/checkpoint-352/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2117dcc2be8117400c495e8c7f1f0db04ff5aee7 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:60fe9e6df17cf0c6681f0df2cf5e28c842e5f132d1abbcfe254459d3c39b3b9f +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..571ea85c4882c8bd75b3fff41181f2cc9597cc43 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:72b936f03af3fa7a3c684bedb1e2325439aed9a3450dbea6497a1eed656d54ac +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..d98bb2e9db712c9e711f32bc1e3142fe2e3e490b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c676299e10f0344839a2b08f1cc1e004410cac0 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2d442c4354d176e51ea6a8521d49ac7a4d4dfa22 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-384/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..983ae2586bcae7fdd029bb39569f4ddfe9f03baf --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.000379653851268813, + "learning_rate": 4.9354244499126866e-05, + "loss": 9.642227269068826e-06, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0017015178455039859, + "learning_rate": 4.90663667927174e-05, + "loss": 1.8409156837151386e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.09125205129384995, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0005971058853901923, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0006, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.006331851705908775, + "learning_rate": 4.849136411748306e-05, + "loss": 6.921300519024953e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00007, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.1789214313030243, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0013293421361595392, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00133, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.63, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.02855750173330307, + "learning_rate": 4.791744483460251e-05, + "loss": 0.00025921204360201955, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00026, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.02957119792699814, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.00019251988851465285, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.024680135771632195, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.00031903525814414024, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00032, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.73, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.1391189694404602, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0013685933081433177, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00137, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.0789211168885231, + "learning_rate": 4.677323837567412e-05, + "loss": 0.0005729947588406503, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00057, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.018962092697620392, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00019800761947408319, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0002, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.06257972121238708, + "learning_rate": 4.620314165804964e-05, + "loss": 0.00047805995563976467, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.13331402838230133, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0005989698111079633, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0006, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.049926936626434326, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0006213528104126453, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00062, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.6875918507575989, + "learning_rate": 4.535077169046201e-05, + "loss": 0.006850574631243944, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00687, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.41269931197166443, + "learning_rate": 4.506743570677743e-05, + "loss": 0.008352026343345642, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00839, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.25145408511161804, + "learning_rate": 4.478451305763618e-05, + "loss": 0.004350180737674236, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00436, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.017037076875567436, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0001216172895510681, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00012, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.049017518758773804, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00043262774124741554, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00043, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.020307866856455803, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00011033139890059829, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0014244886115193367, + "learning_rate": 4.36571910095382e-05, + "loss": 2.7976970159215853e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0005130386562086642, + "learning_rate": 4.337651127546448e-05, + "loss": 1.5001332030806225e-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": 37.59, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.015154760330915451, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00014183670282363892, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00014, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.009964634664356709, + "learning_rate": 4.2816614569850635e-05, + "loss": 7.889936387073249e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00008, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.22056631743907928, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.006305779796093702, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00633, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.024270815774798393, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002287090610479936, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.11968889087438583, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0012203591177240014, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00122, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.28869929909706116, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0029480340890586376, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00295, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.08119015395641327, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0008730962872505188, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00087, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.24371854960918427, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0037464506458491087, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00375, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0005952336359769106, + "learning_rate": 4.087355496656321e-05, + "loss": 1.7879261577036232e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.21333903074264526, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0022262686397880316, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00223, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.12917986512184143, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0012778663076460361, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00128, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.011195745319128036, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00012576431618072093, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.69, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.11680588871240616, + "learning_rate": 3.977591418134619e-05, + "loss": 0.001117704901844263, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00112, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.1339329034090042, + "learning_rate": 3.95030593412612e-05, + "loss": 0.0011167675256729126, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00112, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.00400894321501255, + "learning_rate": 3.923084939213296e-05, + "loss": 5.7869408919941634e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00006, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.012578712776303291, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00020866318664047867, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.4467204213142395, + "learning_rate": 3.868840945050728e-05, + "loss": 0.013730829581618309, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01383, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.002340462524443865, + "learning_rate": 3.841820203114995e-05, + "loss": 4.573522164719179e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.005373048596084118, + "learning_rate": 3.814868464809027e-05, + "loss": 5.34585997229442e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00005, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.06, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.8296674489974976, + "learning_rate": 3.787986851704667e-05, + "loss": 0.009736484847962856, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00978, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.01640630140900612, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00026526901638135314, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.03, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.013102808967232704, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00013475351443048567, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00013, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.01519423071295023, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002597831189632416, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00026, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.002785792574286461, + "learning_rate": 3.68118397962661e-05, + "loss": 5.2896200941177085e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.86, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.045195769518613815, + "learning_rate": 3.654669712344384e-05, + "loss": 0.000801014481112361, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.028165830299258232, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004341424209997058, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00043, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.07202710211277008, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0020600268617272377, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.05288974568247795, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0013123506214469671, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00131, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0019543056841939688, + "learning_rate": 3.549391545931585e-05, + "loss": 4.880658889305778e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.00155684445053339, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.453617293620482e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 30.93, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.012292287312448025, + "learning_rate": 3.49723512647657e-05, + "loss": 0.00024946065968833864, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00025, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.008822723291814327, + "learning_rate": 3.471281389826491e-05, + "loss": 0.00019343552412465215, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.8359783291816711, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.015975475311279297, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0161, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.012664229609072208, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.00016400852473452687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00016, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.01816992089152336, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.00016030078404583037, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.011242903769016266, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.00022866495419293642, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00023, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005438812542706728, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00012295089254621416, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00012, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.53, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.05521470680832863, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0009212898439727724, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00092, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.045007552951574326, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0007900734199211001, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00079, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.18341247737407684, + "learning_rate": 3.266780708475511e-05, + "loss": 0.005364480894058943, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00538, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0726502537727356, + "learning_rate": 3.241625231705354e-05, + "loss": 0.001004248857498169, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.001, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.013726887293159962, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00017499152454547584, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00018, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.15467943251132965, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0034005024936050177, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00341, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.07023883610963821, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0016934729646891356, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00169, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.002377701224759221, + "learning_rate": 3.141953535845912e-05, + "loss": 5.6141776440199465e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00006, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01144740916788578, + "learning_rate": 3.11727834939056e-05, + "loss": 0.00025334605015814304, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00025, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.14478880167007446, + "learning_rate": 3.092702317708967e-05, + "loss": 0.003968201577663422, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00398, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.022581662982702255, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00020885399135295302, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00021, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 2.1134932041168213, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.000105415572761558, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00011, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0530208945274353, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0006053540273569524, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00061, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.032864298671483994, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.00037546915700659156, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00038, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.040343064814805984, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00035573297645896673, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00036, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.009443948976695538, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.00012895476538687944, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00013, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.20513421297073364, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0005669078091159463, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00057, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.005197233986109495, + "learning_rate": 2.8997854750998964e-05, + "loss": 8.792963490122929e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.005207398906350136, + "learning_rate": 2.8761473491751258e-05, + "loss": 7.358018774539232e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00007, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.008117729797959328, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.00011434487532824278, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.52, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.02638891525566578, + "learning_rate": 2.829199644117484e-05, + "loss": 0.00033778423676267266, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00034, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.15596236288547516, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00018454823293723166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00018, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.013470837846398354, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0001927075209096074, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.009815672412514687, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0001360755559289828, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00014, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.003195439698174596, + "learning_rate": 2.7366456756428184e-05, + "loss": 6.715093331877142e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.009018086828291416, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.873787737684324e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.02981177344918251, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0004891970893368125, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00049, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.28394025564193726, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0027279453352093697, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00273, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.06641732156276703, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0007971529266797006, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.75, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.26441285014152527, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0031230500899255276, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00313, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0731891617178917, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0010390954557806253, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.32646894454956055, + "learning_rate": 2.579139666504821e-05, + "loss": 0.011534517630934715, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0116, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.020526761189103127, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0003196616598870605, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00032, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.07209137082099915, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0009562239283695817, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00096, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.015863383188843727, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.00018306101264897734, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00018, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.031188542023301125, + "learning_rate": 2.491789760603361e-05, + "loss": 0.0005925485165789723, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00059, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.03409359231591225, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0006768744206055999, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00068, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.18, + "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_fake_1x__agreement/training/checkpoints/checkpoint-384/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..123f79e57009b549b424830dcd6b8eb6198d9690 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6dcf4d9432a3b232716f6a137541e081bc8dc6634dc3cd8dad0beee00b2c66a +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..b4024a30c47287a593cd98d19f3b6bf00d2c098e --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:48f5aaf54703d16046d39b9a3ab21ce08c21306199f3ee916efaea71c4015f96 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..378ac734ffa403364eeb8e241ac5d35631fe136e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..20a981d8c8f57b8d2e18429ceb299414229ad6ac --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..ca38c2c0f858d125af38a696e17cd1c5dc72b78e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-416/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a278040f692ffd03b8c1423201cf2d91bc74ac2d --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.000379653851268813, + "learning_rate": 4.9354244499126866e-05, + "loss": 9.642227269068826e-06, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0017015178455039859, + "learning_rate": 4.90663667927174e-05, + "loss": 1.8409156837151386e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.09125205129384995, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0005971058853901923, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0006, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.006331851705908775, + "learning_rate": 4.849136411748306e-05, + "loss": 6.921300519024953e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00007, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.1789214313030243, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0013293421361595392, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00133, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.63, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.02855750173330307, + "learning_rate": 4.791744483460251e-05, + "loss": 0.00025921204360201955, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00026, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.02957119792699814, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.00019251988851465285, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.024680135771632195, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.00031903525814414024, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00032, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.73, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.1391189694404602, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0013685933081433177, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00137, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.0789211168885231, + "learning_rate": 4.677323837567412e-05, + "loss": 0.0005729947588406503, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00057, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.018962092697620392, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00019800761947408319, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0002, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.06257972121238708, + "learning_rate": 4.620314165804964e-05, + "loss": 0.00047805995563976467, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.13331402838230133, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0005989698111079633, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0006, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.049926936626434326, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0006213528104126453, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00062, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.6875918507575989, + "learning_rate": 4.535077169046201e-05, + "loss": 0.006850574631243944, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00687, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.41269931197166443, + "learning_rate": 4.506743570677743e-05, + "loss": 0.008352026343345642, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00839, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.25145408511161804, + "learning_rate": 4.478451305763618e-05, + "loss": 0.004350180737674236, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00436, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.017037076875567436, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0001216172895510681, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00012, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.049017518758773804, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00043262774124741554, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00043, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.020307866856455803, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00011033139890059829, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0014244886115193367, + "learning_rate": 4.36571910095382e-05, + "loss": 2.7976970159215853e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0005130386562086642, + "learning_rate": 4.337651127546448e-05, + "loss": 1.5001332030806225e-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": 37.59, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.015154760330915451, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00014183670282363892, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00014, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.009964634664356709, + "learning_rate": 4.2816614569850635e-05, + "loss": 7.889936387073249e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00008, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.22056631743907928, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.006305779796093702, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00633, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.024270815774798393, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002287090610479936, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.11968889087438583, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0012203591177240014, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00122, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.28869929909706116, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0029480340890586376, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00295, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.08119015395641327, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0008730962872505188, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00087, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.24371854960918427, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0037464506458491087, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00375, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0005952336359769106, + "learning_rate": 4.087355496656321e-05, + "loss": 1.7879261577036232e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.21333903074264526, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0022262686397880316, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00223, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.12917986512184143, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0012778663076460361, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00128, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.011195745319128036, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00012576431618072093, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.69, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.11680588871240616, + "learning_rate": 3.977591418134619e-05, + "loss": 0.001117704901844263, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00112, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.1339329034090042, + "learning_rate": 3.95030593412612e-05, + "loss": 0.0011167675256729126, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00112, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.00400894321501255, + "learning_rate": 3.923084939213296e-05, + "loss": 5.7869408919941634e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00006, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.012578712776303291, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00020866318664047867, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.4467204213142395, + "learning_rate": 3.868840945050728e-05, + "loss": 0.013730829581618309, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01383, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.002340462524443865, + "learning_rate": 3.841820203114995e-05, + "loss": 4.573522164719179e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.005373048596084118, + "learning_rate": 3.814868464809027e-05, + "loss": 5.34585997229442e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00005, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.06, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.8296674489974976, + "learning_rate": 3.787986851704667e-05, + "loss": 0.009736484847962856, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00978, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.01640630140900612, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00026526901638135314, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.03, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.013102808967232704, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00013475351443048567, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00013, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.01519423071295023, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002597831189632416, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00026, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.002785792574286461, + "learning_rate": 3.68118397962661e-05, + "loss": 5.2896200941177085e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.86, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.045195769518613815, + "learning_rate": 3.654669712344384e-05, + "loss": 0.000801014481112361, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.028165830299258232, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004341424209997058, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00043, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.07202710211277008, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0020600268617272377, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.05288974568247795, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0013123506214469671, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00131, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0019543056841939688, + "learning_rate": 3.549391545931585e-05, + "loss": 4.880658889305778e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.00155684445053339, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.453617293620482e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 30.93, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.012292287312448025, + "learning_rate": 3.49723512647657e-05, + "loss": 0.00024946065968833864, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00025, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.008822723291814327, + "learning_rate": 3.471281389826491e-05, + "loss": 0.00019343552412465215, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.8359783291816711, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.015975475311279297, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0161, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.012664229609072208, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.00016400852473452687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00016, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.01816992089152336, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.00016030078404583037, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.011242903769016266, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.00022866495419293642, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00023, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005438812542706728, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00012295089254621416, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00012, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.53, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.05521470680832863, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0009212898439727724, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00092, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.045007552951574326, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0007900734199211001, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00079, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.18341247737407684, + "learning_rate": 3.266780708475511e-05, + "loss": 0.005364480894058943, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00538, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0726502537727356, + "learning_rate": 3.241625231705354e-05, + "loss": 0.001004248857498169, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.001, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.013726887293159962, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00017499152454547584, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00018, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.15467943251132965, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0034005024936050177, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00341, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.07023883610963821, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0016934729646891356, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00169, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.002377701224759221, + "learning_rate": 3.141953535845912e-05, + "loss": 5.6141776440199465e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00006, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01144740916788578, + "learning_rate": 3.11727834939056e-05, + "loss": 0.00025334605015814304, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00025, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.14478880167007446, + "learning_rate": 3.092702317708967e-05, + "loss": 0.003968201577663422, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00398, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.022581662982702255, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00020885399135295302, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00021, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 2.1134932041168213, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.000105415572761558, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00011, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0530208945274353, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0006053540273569524, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00061, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.032864298671483994, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.00037546915700659156, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00038, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.040343064814805984, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00035573297645896673, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00036, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.009443948976695538, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.00012895476538687944, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00013, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.20513421297073364, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0005669078091159463, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00057, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.005197233986109495, + "learning_rate": 2.8997854750998964e-05, + "loss": 8.792963490122929e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.005207398906350136, + "learning_rate": 2.8761473491751258e-05, + "loss": 7.358018774539232e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00007, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.008117729797959328, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.00011434487532824278, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.52, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.02638891525566578, + "learning_rate": 2.829199644117484e-05, + "loss": 0.00033778423676267266, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00034, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.15596236288547516, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00018454823293723166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00018, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.013470837846398354, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0001927075209096074, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.009815672412514687, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0001360755559289828, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00014, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.003195439698174596, + "learning_rate": 2.7366456756428184e-05, + "loss": 6.715093331877142e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.009018086828291416, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.873787737684324e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.02981177344918251, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0004891970893368125, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00049, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.28394025564193726, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0027279453352093697, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00273, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.06641732156276703, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0007971529266797006, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.75, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.26441285014152527, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0031230500899255276, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00313, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0731891617178917, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0010390954557806253, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.32646894454956055, + "learning_rate": 2.579139666504821e-05, + "loss": 0.011534517630934715, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0116, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.020526761189103127, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0003196616598870605, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00032, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.07209137082099915, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0009562239283695817, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00096, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.015863383188843727, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.00018306101264897734, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00018, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.031188542023301125, + "learning_rate": 2.491789760603361e-05, + "loss": 0.0005925485165789723, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00059, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.03409359231591225, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0006768744206055999, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00068, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.022114347666502, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.00034674766357056797, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00035, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.007172535173594952, + "learning_rate": 2.427588563158384e-05, + "loss": 0.00010547737474553287, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00011, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.011641777120530605, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00017262669280171394, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00017, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.0061674294993281364, + "learning_rate": 2.3854255584458547e-05, + "loss": 9.34262279770337e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.04236450046300888, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0003042269090656191, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0003, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.013182509690523148, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.00026280959718860686, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00026, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.012172302231192589, + "learning_rate": 2.3231552870624487e-05, + "loss": 8.891556353773922e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.09926214069128036, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.0020971111953258514, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0021, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.006282840855419636, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00012446771143004298, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00012, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.72, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.012727927416563034, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0001671911304583773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00017, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.007153753656893969, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.00010278529953211546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0001, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.026691973209381104, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00036194472340866923, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.1592523753643036, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0023322016932070255, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00233, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.02613229863345623, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0003333169443067163, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00033, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.006405193824321032, + "learning_rate": 2.1629798596457056e-05, + "loss": 8.373583841603249e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00008, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.01651483029127121, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.0001667532924329862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.0015016240067780018, + "learning_rate": 2.124308220868431e-05, + "loss": 3.507245855871588e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 32.89, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.03716011717915535, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0006280053639784455, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00063, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.0026928475126624107, + "learning_rate": 2.0861984815011552e-05, + "loss": 6.0345602832967415e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.011999940499663353, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010794817353598773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.003504133317619562, + "learning_rate": 2.0486569850851317e-05, + "loss": 4.769703446072526e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0031612543389201164, + "learning_rate": 2.0301012845054956e-05, + "loss": 5.125769166625105e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.66, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.005299390759319067, + "learning_rate": 2.011689980574966e-05, + "loss": 7.910770364105701e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00008, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.003536214353516698, + "learning_rate": 1.993423839463052e-05, + "loss": 6.525318895000964e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011145357973873615, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00013281722203828394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00013, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.004464628174901009, + "learning_rate": 1.957330080137385e-05, + "loss": 8.551983773941174e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.004235507920384407, + "learning_rate": 1.9395039639322864e-05, + "loss": 7.31897889636457e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00007, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.0010779709555208683, + "learning_rate": 1.9218260145006073e-05, + "loss": 1.925879405462183e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.043327320367097855, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0004905068781226873, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00049, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.13874047994613647, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003094755229540169, + "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": 36.7, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0014462472172454, + "learning_rate": 1.869688492349885e-05, + "loss": 3.0223447538446635e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.0014046201249584556, + "learning_rate": 1.85261050441233e-05, + "loss": 2.9643942980328575e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.86, + "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_fake_1x__agreement/training/checkpoints/checkpoint-416/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..6dbf7848f55de1c97671bee0bba7913b132afcd9 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:61a60a92d86b34d1299dad4ca5e0fe972f8e3e25b11df37f289810264aada090 +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..58a5a4064f31935d42b79fb323541704c9f84bad --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6575f6ad91d9d213255a3cf0441c2fb611bd35649fa1343de06c7d2e2a1619dc +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..1329b266df9f032a7779db44cbe92f7110ae5a5a --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2045249a1135f9f2d87c87c3e02e6227697d0cfa --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fa78a58752cdd4bcff72584e7054dccc1d7ab6d6 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-448/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3b1bdaf3b640e55f1bc1dd7aff945cd0e3187e7e --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.000379653851268813, + "learning_rate": 4.9354244499126866e-05, + "loss": 9.642227269068826e-06, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0017015178455039859, + "learning_rate": 4.90663667927174e-05, + "loss": 1.8409156837151386e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.09125205129384995, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0005971058853901923, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0006, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.006331851705908775, + "learning_rate": 4.849136411748306e-05, + "loss": 6.921300519024953e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00007, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.1789214313030243, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0013293421361595392, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00133, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.63, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.02855750173330307, + "learning_rate": 4.791744483460251e-05, + "loss": 0.00025921204360201955, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00026, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.02957119792699814, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.00019251988851465285, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.024680135771632195, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.00031903525814414024, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00032, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.73, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.1391189694404602, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0013685933081433177, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00137, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.0789211168885231, + "learning_rate": 4.677323837567412e-05, + "loss": 0.0005729947588406503, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00057, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.018962092697620392, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00019800761947408319, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0002, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.06257972121238708, + "learning_rate": 4.620314165804964e-05, + "loss": 0.00047805995563976467, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.13331402838230133, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0005989698111079633, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0006, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.049926936626434326, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0006213528104126453, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00062, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.6875918507575989, + "learning_rate": 4.535077169046201e-05, + "loss": 0.006850574631243944, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00687, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.41269931197166443, + "learning_rate": 4.506743570677743e-05, + "loss": 0.008352026343345642, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00839, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.25145408511161804, + "learning_rate": 4.478451305763618e-05, + "loss": 0.004350180737674236, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00436, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.017037076875567436, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0001216172895510681, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00012, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.049017518758773804, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00043262774124741554, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00043, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.020307866856455803, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00011033139890059829, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0014244886115193367, + "learning_rate": 4.36571910095382e-05, + "loss": 2.7976970159215853e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0005130386562086642, + "learning_rate": 4.337651127546448e-05, + "loss": 1.5001332030806225e-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": 37.59, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.015154760330915451, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00014183670282363892, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00014, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.009964634664356709, + "learning_rate": 4.2816614569850635e-05, + "loss": 7.889936387073249e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00008, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.22056631743907928, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.006305779796093702, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00633, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.024270815774798393, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002287090610479936, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.11968889087438583, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0012203591177240014, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00122, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.28869929909706116, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0029480340890586376, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00295, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.08119015395641327, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0008730962872505188, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00087, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.24371854960918427, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0037464506458491087, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00375, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0005952336359769106, + "learning_rate": 4.087355496656321e-05, + "loss": 1.7879261577036232e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.21333903074264526, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0022262686397880316, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00223, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.12917986512184143, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0012778663076460361, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00128, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.011195745319128036, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00012576431618072093, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.69, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.11680588871240616, + "learning_rate": 3.977591418134619e-05, + "loss": 0.001117704901844263, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00112, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.1339329034090042, + "learning_rate": 3.95030593412612e-05, + "loss": 0.0011167675256729126, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00112, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.00400894321501255, + "learning_rate": 3.923084939213296e-05, + "loss": 5.7869408919941634e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00006, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.012578712776303291, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00020866318664047867, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.4467204213142395, + "learning_rate": 3.868840945050728e-05, + "loss": 0.013730829581618309, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01383, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.002340462524443865, + "learning_rate": 3.841820203114995e-05, + "loss": 4.573522164719179e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.005373048596084118, + "learning_rate": 3.814868464809027e-05, + "loss": 5.34585997229442e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00005, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.06, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.8296674489974976, + "learning_rate": 3.787986851704667e-05, + "loss": 0.009736484847962856, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00978, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.01640630140900612, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00026526901638135314, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.03, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.013102808967232704, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00013475351443048567, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00013, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.01519423071295023, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002597831189632416, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00026, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.002785792574286461, + "learning_rate": 3.68118397962661e-05, + "loss": 5.2896200941177085e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.86, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.045195769518613815, + "learning_rate": 3.654669712344384e-05, + "loss": 0.000801014481112361, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.028165830299258232, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004341424209997058, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00043, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.07202710211277008, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0020600268617272377, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.05288974568247795, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0013123506214469671, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00131, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0019543056841939688, + "learning_rate": 3.549391545931585e-05, + "loss": 4.880658889305778e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.00155684445053339, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.453617293620482e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 30.93, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.012292287312448025, + "learning_rate": 3.49723512647657e-05, + "loss": 0.00024946065968833864, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00025, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.008822723291814327, + "learning_rate": 3.471281389826491e-05, + "loss": 0.00019343552412465215, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.8359783291816711, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.015975475311279297, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0161, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.012664229609072208, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.00016400852473452687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00016, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.01816992089152336, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.00016030078404583037, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.011242903769016266, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.00022866495419293642, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00023, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005438812542706728, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00012295089254621416, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00012, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.53, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.05521470680832863, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0009212898439727724, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00092, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.045007552951574326, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0007900734199211001, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00079, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.18341247737407684, + "learning_rate": 3.266780708475511e-05, + "loss": 0.005364480894058943, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00538, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0726502537727356, + "learning_rate": 3.241625231705354e-05, + "loss": 0.001004248857498169, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.001, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.013726887293159962, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00017499152454547584, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00018, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.15467943251132965, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0034005024936050177, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00341, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.07023883610963821, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0016934729646891356, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00169, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.002377701224759221, + "learning_rate": 3.141953535845912e-05, + "loss": 5.6141776440199465e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00006, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01144740916788578, + "learning_rate": 3.11727834939056e-05, + "loss": 0.00025334605015814304, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00025, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.14478880167007446, + "learning_rate": 3.092702317708967e-05, + "loss": 0.003968201577663422, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00398, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.022581662982702255, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00020885399135295302, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00021, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 2.1134932041168213, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.000105415572761558, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00011, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0530208945274353, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0006053540273569524, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00061, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.032864298671483994, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.00037546915700659156, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00038, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.040343064814805984, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00035573297645896673, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00036, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.009443948976695538, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.00012895476538687944, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00013, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.20513421297073364, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0005669078091159463, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00057, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.005197233986109495, + "learning_rate": 2.8997854750998964e-05, + "loss": 8.792963490122929e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.005207398906350136, + "learning_rate": 2.8761473491751258e-05, + "loss": 7.358018774539232e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00007, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.008117729797959328, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.00011434487532824278, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.52, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.02638891525566578, + "learning_rate": 2.829199644117484e-05, + "loss": 0.00033778423676267266, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00034, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.15596236288547516, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00018454823293723166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00018, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.013470837846398354, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0001927075209096074, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.009815672412514687, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0001360755559289828, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00014, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.003195439698174596, + "learning_rate": 2.7366456756428184e-05, + "loss": 6.715093331877142e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.009018086828291416, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.873787737684324e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.02981177344918251, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0004891970893368125, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00049, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.28394025564193726, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0027279453352093697, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00273, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.06641732156276703, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0007971529266797006, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.75, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.26441285014152527, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0031230500899255276, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00313, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0731891617178917, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0010390954557806253, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.32646894454956055, + "learning_rate": 2.579139666504821e-05, + "loss": 0.011534517630934715, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0116, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.020526761189103127, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0003196616598870605, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00032, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.07209137082099915, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0009562239283695817, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00096, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.015863383188843727, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.00018306101264897734, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00018, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.031188542023301125, + "learning_rate": 2.491789760603361e-05, + "loss": 0.0005925485165789723, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00059, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.03409359231591225, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0006768744206055999, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00068, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.022114347666502, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.00034674766357056797, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00035, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.007172535173594952, + "learning_rate": 2.427588563158384e-05, + "loss": 0.00010547737474553287, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00011, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.011641777120530605, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00017262669280171394, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00017, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.0061674294993281364, + "learning_rate": 2.3854255584458547e-05, + "loss": 9.34262279770337e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.04236450046300888, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0003042269090656191, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0003, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.013182509690523148, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.00026280959718860686, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00026, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.012172302231192589, + "learning_rate": 2.3231552870624487e-05, + "loss": 8.891556353773922e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.09926214069128036, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.0020971111953258514, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0021, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.006282840855419636, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00012446771143004298, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00012, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.72, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.012727927416563034, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0001671911304583773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00017, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.007153753656893969, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.00010278529953211546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0001, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.026691973209381104, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00036194472340866923, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.1592523753643036, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0023322016932070255, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00233, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.02613229863345623, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0003333169443067163, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00033, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.006405193824321032, + "learning_rate": 2.1629798596457056e-05, + "loss": 8.373583841603249e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00008, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.01651483029127121, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.0001667532924329862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.0015016240067780018, + "learning_rate": 2.124308220868431e-05, + "loss": 3.507245855871588e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 32.89, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.03716011717915535, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0006280053639784455, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00063, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.0026928475126624107, + "learning_rate": 2.0861984815011552e-05, + "loss": 6.0345602832967415e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.011999940499663353, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010794817353598773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.003504133317619562, + "learning_rate": 2.0486569850851317e-05, + "loss": 4.769703446072526e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0031612543389201164, + "learning_rate": 2.0301012845054956e-05, + "loss": 5.125769166625105e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.66, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.005299390759319067, + "learning_rate": 2.011689980574966e-05, + "loss": 7.910770364105701e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00008, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.003536214353516698, + "learning_rate": 1.993423839463052e-05, + "loss": 6.525318895000964e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011145357973873615, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00013281722203828394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00013, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.004464628174901009, + "learning_rate": 1.957330080137385e-05, + "loss": 8.551983773941174e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.004235507920384407, + "learning_rate": 1.9395039639322864e-05, + "loss": 7.31897889636457e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00007, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.0010779709555208683, + "learning_rate": 1.9218260145006073e-05, + "loss": 1.925879405462183e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.043327320367097855, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0004905068781226873, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00049, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.13874047994613647, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003094755229540169, + "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": 36.7, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0014462472172454, + "learning_rate": 1.869688492349885e-05, + "loss": 3.0223447538446635e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.0014046201249584556, + "learning_rate": 1.85261050441233e-05, + "loss": 2.9643942980328575e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.000552138953935355, + "learning_rate": 1.8356842992397304e-05, + "loss": 1.5713212633272633e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.005376334302127361, + "learning_rate": 1.8189105812005714e-05, + "loss": 6.261414819164202e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00006, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.00120884261559695, + "learning_rate": 1.802290048317732e-05, + "loss": 2.686155858100392e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.004546779207885265, + "learning_rate": 1.785823392239424e-05, + "loss": 7.581210957141593e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00008, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.014719134196639061, + "learning_rate": 1.7695112982104225e-05, + "loss": 0.00021209896658547223, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00021, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.0006049374933354557, + "learning_rate": 1.7533544450435433e-05, + "loss": 1.7317186575382948e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00002, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.01476835086941719, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.0001859527692431584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00019, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.05, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.003950290847569704, + "learning_rate": 1.721509144218405e-05, + "loss": 9.087299986276776e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.008329745382070541, + "learning_rate": 1.705822021773101e-05, + "loss": 8.313520083902404e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.5991763472557068, + "learning_rate": 1.69029279056068e-05, + "loss": 0.004271846264600754, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00428, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.02893899939954281, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.00015647552208974957, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0829748809337616, + "learning_rate": 1.659710580175893e-05, + "loss": 0.0008744620135985315, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.00444152532145381, + "learning_rate": 1.644658873654133e-05, + "loss": 6.830539496149868e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 429, + "tokens/total": 12985328, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.02823544666171074, + "learning_rate": 1.629767603613508e-05, + "loss": 0.0002984766324516386, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0003, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.76, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 6.060963153839111, + "learning_rate": 1.615037389740547e-05, + "loss": 0.00036459555849432945, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00036, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.12, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.01055138185620308, + "learning_rate": 1.600468845019576e-05, + "loss": 0.00012494996190071106, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00012, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.00017622209270484746, + "learning_rate": 1.5860625757072092e-05, + "loss": 6.37679568171734e-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": 31.6, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.04205000028014183, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00029261858435347676, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00029, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.012470721267163754, + "learning_rate": 1.557739254545075e-05, + "loss": 0.00017825230315793306, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.002119631739333272, + "learning_rate": 1.543823381344311e-05, + "loss": 3.121166810160503e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00003, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.005058391951024532, + "learning_rate": 1.5300721408011114e-05, + "loss": 5.037401570007205e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.0023086639121174812, + "learning_rate": 1.5164861051607254e-05, + "loss": 3.312892295070924e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.0005081371637061238, + "learning_rate": 1.5030658397935521e-05, + "loss": 1.3808794392389245e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.03984901309013367, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00027867406606674194, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00028, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.004065736196935177, + "learning_rate": 1.476724846845306e-05, + "loss": 4.005522350780666e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 441, + "tokens/total": 13349392, + "tokens/train_per_sec_per_gpu": 40.07, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.0009828282054513693, + "learning_rate": 1.463805215420471e-05, + "loss": 2.004339148697909e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.0016327777411788702, + "learning_rate": 1.451053546535705e-05, + "loss": 2.840825436578598e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.07271858304738998, + "learning_rate": 1.438470370840001e-05, + "loss": 0.00016209090244956315, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00016, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.0013328593922778964, + "learning_rate": 1.4260562119706606e-05, + "loss": 2.0676230633398518e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00002, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.34, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.0807131826877594, + "learning_rate": 1.413811586531508e-05, + "loss": 0.0006098271114751697, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00061, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.010202539153397083, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.00011388547136448324, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00011, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.012697691097855568, + "learning_rate": 1.3898329670629645e-05, + "loss": 6.711178866680712e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.23, + "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_fake_1x__agreement/training/checkpoints/checkpoint-448/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..07314b330a575c1f4b00843bbfeb3c48c5156a23 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2846d936ab50326f9fa67f1a7b9a26399c44250a961524ef11cb53b6221187b0 +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..f722dfc93920d9364b1ac98a8f342e93d018fae5 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39e28ab60da1dd67f54787c0c1f918eedf463228be2e1402cbe65e78d6e8b5de +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..9da896c9d8f84289d77b53501f8d2b2d3e8bc3ec --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce1c4e63a6c6f8be8c137d3add4eda184d59a043 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3da71c55273035b5e1d531117df684eced67a866 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-480/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..53bba44021aa35014ddcb08b347ebfb8fa2793c4 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.000379653851268813, + "learning_rate": 4.9354244499126866e-05, + "loss": 9.642227269068826e-06, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0017015178455039859, + "learning_rate": 4.90663667927174e-05, + "loss": 1.8409156837151386e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.09125205129384995, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0005971058853901923, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0006, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.006331851705908775, + "learning_rate": 4.849136411748306e-05, + "loss": 6.921300519024953e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00007, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.1789214313030243, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0013293421361595392, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00133, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.63, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.02855750173330307, + "learning_rate": 4.791744483460251e-05, + "loss": 0.00025921204360201955, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00026, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.02957119792699814, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.00019251988851465285, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.024680135771632195, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.00031903525814414024, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00032, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.73, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.1391189694404602, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0013685933081433177, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00137, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.0789211168885231, + "learning_rate": 4.677323837567412e-05, + "loss": 0.0005729947588406503, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00057, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.018962092697620392, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00019800761947408319, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0002, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.06257972121238708, + "learning_rate": 4.620314165804964e-05, + "loss": 0.00047805995563976467, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.13331402838230133, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0005989698111079633, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0006, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.049926936626434326, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0006213528104126453, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00062, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.6875918507575989, + "learning_rate": 4.535077169046201e-05, + "loss": 0.006850574631243944, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00687, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.41269931197166443, + "learning_rate": 4.506743570677743e-05, + "loss": 0.008352026343345642, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00839, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.25145408511161804, + "learning_rate": 4.478451305763618e-05, + "loss": 0.004350180737674236, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00436, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.017037076875567436, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0001216172895510681, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00012, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.049017518758773804, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00043262774124741554, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00043, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.020307866856455803, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00011033139890059829, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0014244886115193367, + "learning_rate": 4.36571910095382e-05, + "loss": 2.7976970159215853e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0005130386562086642, + "learning_rate": 4.337651127546448e-05, + "loss": 1.5001332030806225e-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": 37.59, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.015154760330915451, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00014183670282363892, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00014, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.009964634664356709, + "learning_rate": 4.2816614569850635e-05, + "loss": 7.889936387073249e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00008, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.22056631743907928, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.006305779796093702, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00633, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.024270815774798393, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002287090610479936, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.11968889087438583, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0012203591177240014, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00122, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.28869929909706116, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0029480340890586376, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00295, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.08119015395641327, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0008730962872505188, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00087, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.24371854960918427, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0037464506458491087, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00375, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0005952336359769106, + "learning_rate": 4.087355496656321e-05, + "loss": 1.7879261577036232e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.21333903074264526, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0022262686397880316, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00223, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.12917986512184143, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0012778663076460361, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00128, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.011195745319128036, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00012576431618072093, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.69, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.11680588871240616, + "learning_rate": 3.977591418134619e-05, + "loss": 0.001117704901844263, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00112, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.1339329034090042, + "learning_rate": 3.95030593412612e-05, + "loss": 0.0011167675256729126, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00112, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.00400894321501255, + "learning_rate": 3.923084939213296e-05, + "loss": 5.7869408919941634e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00006, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.012578712776303291, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00020866318664047867, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.4467204213142395, + "learning_rate": 3.868840945050728e-05, + "loss": 0.013730829581618309, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01383, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.002340462524443865, + "learning_rate": 3.841820203114995e-05, + "loss": 4.573522164719179e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.005373048596084118, + "learning_rate": 3.814868464809027e-05, + "loss": 5.34585997229442e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00005, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.06, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.8296674489974976, + "learning_rate": 3.787986851704667e-05, + "loss": 0.009736484847962856, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00978, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.01640630140900612, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00026526901638135314, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.03, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.013102808967232704, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00013475351443048567, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00013, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.01519423071295023, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002597831189632416, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00026, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.002785792574286461, + "learning_rate": 3.68118397962661e-05, + "loss": 5.2896200941177085e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.86, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.045195769518613815, + "learning_rate": 3.654669712344384e-05, + "loss": 0.000801014481112361, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.028165830299258232, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004341424209997058, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00043, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.07202710211277008, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0020600268617272377, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.05288974568247795, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0013123506214469671, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00131, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0019543056841939688, + "learning_rate": 3.549391545931585e-05, + "loss": 4.880658889305778e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.00155684445053339, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.453617293620482e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 30.93, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.012292287312448025, + "learning_rate": 3.49723512647657e-05, + "loss": 0.00024946065968833864, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00025, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.008822723291814327, + "learning_rate": 3.471281389826491e-05, + "loss": 0.00019343552412465215, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.8359783291816711, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.015975475311279297, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0161, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.012664229609072208, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.00016400852473452687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00016, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.01816992089152336, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.00016030078404583037, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.011242903769016266, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.00022866495419293642, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00023, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005438812542706728, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00012295089254621416, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00012, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.53, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.05521470680832863, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0009212898439727724, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00092, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.045007552951574326, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0007900734199211001, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00079, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.18341247737407684, + "learning_rate": 3.266780708475511e-05, + "loss": 0.005364480894058943, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00538, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0726502537727356, + "learning_rate": 3.241625231705354e-05, + "loss": 0.001004248857498169, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.001, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.013726887293159962, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00017499152454547584, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00018, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.15467943251132965, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0034005024936050177, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00341, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.07023883610963821, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0016934729646891356, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00169, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.002377701224759221, + "learning_rate": 3.141953535845912e-05, + "loss": 5.6141776440199465e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00006, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01144740916788578, + "learning_rate": 3.11727834939056e-05, + "loss": 0.00025334605015814304, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00025, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.14478880167007446, + "learning_rate": 3.092702317708967e-05, + "loss": 0.003968201577663422, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00398, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.022581662982702255, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00020885399135295302, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00021, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 2.1134932041168213, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.000105415572761558, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00011, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0530208945274353, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0006053540273569524, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00061, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.032864298671483994, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.00037546915700659156, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00038, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.040343064814805984, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00035573297645896673, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00036, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.009443948976695538, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.00012895476538687944, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00013, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.20513421297073364, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0005669078091159463, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00057, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.005197233986109495, + "learning_rate": 2.8997854750998964e-05, + "loss": 8.792963490122929e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.005207398906350136, + "learning_rate": 2.8761473491751258e-05, + "loss": 7.358018774539232e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00007, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.008117729797959328, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.00011434487532824278, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.52, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.02638891525566578, + "learning_rate": 2.829199644117484e-05, + "loss": 0.00033778423676267266, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00034, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.15596236288547516, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00018454823293723166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00018, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.013470837846398354, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0001927075209096074, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.009815672412514687, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0001360755559289828, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00014, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.003195439698174596, + "learning_rate": 2.7366456756428184e-05, + "loss": 6.715093331877142e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.009018086828291416, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.873787737684324e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.02981177344918251, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0004891970893368125, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00049, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.28394025564193726, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0027279453352093697, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00273, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.06641732156276703, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0007971529266797006, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.75, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.26441285014152527, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0031230500899255276, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00313, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0731891617178917, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0010390954557806253, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.32646894454956055, + "learning_rate": 2.579139666504821e-05, + "loss": 0.011534517630934715, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0116, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.020526761189103127, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0003196616598870605, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00032, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.07209137082099915, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0009562239283695817, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00096, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.015863383188843727, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.00018306101264897734, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00018, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.031188542023301125, + "learning_rate": 2.491789760603361e-05, + "loss": 0.0005925485165789723, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00059, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.03409359231591225, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0006768744206055999, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00068, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.022114347666502, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.00034674766357056797, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00035, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.007172535173594952, + "learning_rate": 2.427588563158384e-05, + "loss": 0.00010547737474553287, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00011, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.011641777120530605, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00017262669280171394, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00017, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.0061674294993281364, + "learning_rate": 2.3854255584458547e-05, + "loss": 9.34262279770337e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.04236450046300888, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0003042269090656191, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0003, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.013182509690523148, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.00026280959718860686, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00026, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.012172302231192589, + "learning_rate": 2.3231552870624487e-05, + "loss": 8.891556353773922e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.09926214069128036, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.0020971111953258514, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0021, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.006282840855419636, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00012446771143004298, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00012, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.72, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.012727927416563034, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0001671911304583773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00017, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.007153753656893969, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.00010278529953211546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0001, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.026691973209381104, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00036194472340866923, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.1592523753643036, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0023322016932070255, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00233, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.02613229863345623, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0003333169443067163, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00033, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.006405193824321032, + "learning_rate": 2.1629798596457056e-05, + "loss": 8.373583841603249e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00008, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.01651483029127121, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.0001667532924329862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.0015016240067780018, + "learning_rate": 2.124308220868431e-05, + "loss": 3.507245855871588e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 32.89, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.03716011717915535, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0006280053639784455, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00063, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.0026928475126624107, + "learning_rate": 2.0861984815011552e-05, + "loss": 6.0345602832967415e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.011999940499663353, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010794817353598773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.003504133317619562, + "learning_rate": 2.0486569850851317e-05, + "loss": 4.769703446072526e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0031612543389201164, + "learning_rate": 2.0301012845054956e-05, + "loss": 5.125769166625105e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.66, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.005299390759319067, + "learning_rate": 2.011689980574966e-05, + "loss": 7.910770364105701e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00008, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.003536214353516698, + "learning_rate": 1.993423839463052e-05, + "loss": 6.525318895000964e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011145357973873615, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00013281722203828394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00013, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.004464628174901009, + "learning_rate": 1.957330080137385e-05, + "loss": 8.551983773941174e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.004235507920384407, + "learning_rate": 1.9395039639322864e-05, + "loss": 7.31897889636457e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00007, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.0010779709555208683, + "learning_rate": 1.9218260145006073e-05, + "loss": 1.925879405462183e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.043327320367097855, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0004905068781226873, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00049, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.13874047994613647, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003094755229540169, + "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": 36.7, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0014462472172454, + "learning_rate": 1.869688492349885e-05, + "loss": 3.0223447538446635e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.0014046201249584556, + "learning_rate": 1.85261050441233e-05, + "loss": 2.9643942980328575e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.000552138953935355, + "learning_rate": 1.8356842992397304e-05, + "loss": 1.5713212633272633e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.005376334302127361, + "learning_rate": 1.8189105812005714e-05, + "loss": 6.261414819164202e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00006, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.00120884261559695, + "learning_rate": 1.802290048317732e-05, + "loss": 2.686155858100392e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.004546779207885265, + "learning_rate": 1.785823392239424e-05, + "loss": 7.581210957141593e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00008, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.014719134196639061, + "learning_rate": 1.7695112982104225e-05, + "loss": 0.00021209896658547223, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00021, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.0006049374933354557, + "learning_rate": 1.7533544450435433e-05, + "loss": 1.7317186575382948e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00002, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.01476835086941719, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.0001859527692431584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00019, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.05, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.003950290847569704, + "learning_rate": 1.721509144218405e-05, + "loss": 9.087299986276776e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.008329745382070541, + "learning_rate": 1.705822021773101e-05, + "loss": 8.313520083902404e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.5991763472557068, + "learning_rate": 1.69029279056068e-05, + "loss": 0.004271846264600754, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00428, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.02893899939954281, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.00015647552208974957, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0829748809337616, + "learning_rate": 1.659710580175893e-05, + "loss": 0.0008744620135985315, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.00444152532145381, + "learning_rate": 1.644658873654133e-05, + "loss": 6.830539496149868e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 429, + "tokens/total": 12985328, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.02823544666171074, + "learning_rate": 1.629767603613508e-05, + "loss": 0.0002984766324516386, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0003, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.76, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 6.060963153839111, + "learning_rate": 1.615037389740547e-05, + "loss": 0.00036459555849432945, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00036, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.12, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.01055138185620308, + "learning_rate": 1.600468845019576e-05, + "loss": 0.00012494996190071106, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00012, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.00017622209270484746, + "learning_rate": 1.5860625757072092e-05, + "loss": 6.37679568171734e-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": 31.6, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.04205000028014183, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00029261858435347676, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00029, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.012470721267163754, + "learning_rate": 1.557739254545075e-05, + "loss": 0.00017825230315793306, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.002119631739333272, + "learning_rate": 1.543823381344311e-05, + "loss": 3.121166810160503e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00003, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.005058391951024532, + "learning_rate": 1.5300721408011114e-05, + "loss": 5.037401570007205e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.0023086639121174812, + "learning_rate": 1.5164861051607254e-05, + "loss": 3.312892295070924e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.0005081371637061238, + "learning_rate": 1.5030658397935521e-05, + "loss": 1.3808794392389245e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.03984901309013367, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00027867406606674194, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00028, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.004065736196935177, + "learning_rate": 1.476724846845306e-05, + "loss": 4.005522350780666e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 441, + "tokens/total": 13349392, + "tokens/train_per_sec_per_gpu": 40.07, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.0009828282054513693, + "learning_rate": 1.463805215420471e-05, + "loss": 2.004339148697909e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.0016327777411788702, + "learning_rate": 1.451053546535705e-05, + "loss": 2.840825436578598e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.07271858304738998, + "learning_rate": 1.438470370840001e-05, + "loss": 0.00016209090244956315, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00016, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.0013328593922778964, + "learning_rate": 1.4260562119706606e-05, + "loss": 2.0676230633398518e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00002, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.34, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.0807131826877594, + "learning_rate": 1.413811586531508e-05, + "loss": 0.0006098271114751697, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00061, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.010202539153397083, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.00011388547136448324, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00011, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.012697691097855568, + "learning_rate": 1.3898329670629645e-05, + "loss": 6.711178866680712e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.23, + "tokens/trainable": 204821 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.00029595353407785296, + "learning_rate": 1.3780999708818058e-05, + "loss": 8.797971531748772e-06, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 449, + "tokens/total": 13590496, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 205271 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.022445866838097572, + "learning_rate": 1.3665385037857758e-05, + "loss": 0.00015267175331246108, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00015, + "step": 450, + "tokens/total": 13620960, + "tokens/train_per_sec_per_gpu": 31.15, + "tokens/trainable": 205723 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.0014366039540618658, + "learning_rate": 1.3551490468947126e-05, + "loss": 2.87473012576811e-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": 32.65, + "tokens/trainable": 206169 + }, + { + "epoch": 1.765625, + "grad_norm": 0.0042471266351640224, + "learning_rate": 1.3439320741704075e-05, + "loss": 5.8180645282845944e-05, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00006, + "step": 452, + "tokens/total": 13681792, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 206592 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.009985425509512424, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.0001178122911369428, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00012, + "step": 453, + "tokens/total": 13712256, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 207055 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.024878855794668198, + "learning_rate": 1.3220174411609587e-05, + "loss": 0.00022001308389008045, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00022, + "step": 454, + "tokens/total": 13742128, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 207494 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.002308261813595891, + "learning_rate": 1.3113206928331471e-05, + "loss": 2.7343181500327773e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 455, + "tokens/total": 13772160, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 207956 + }, + { + "epoch": 1.78125, + "grad_norm": 0.0006800597766414285, + "learning_rate": 1.300798252548806e-05, + "loss": 1.6906453311094083e-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": 35.72, + "tokens/trainable": 208408 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.04438555985689163, + "learning_rate": 1.2904505581896265e-05, + "loss": 0.0004473269509617239, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00045, + "step": 457, + "tokens/total": 13832880, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 208889 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.04841502383351326, + "learning_rate": 1.2802780403654082e-05, + "loss": 0.00028527071117423475, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00029, + "step": 458, + "tokens/total": 13862880, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 209336 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.011998557485640049, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.00014353993174154311, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 459, + "tokens/total": 13893440, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 209797 + }, + { + "epoch": 1.796875, + "grad_norm": 0.0004317575367167592, + "learning_rate": 1.2604602202943861e-05, + "loss": 1.386312760587316e-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": 36.34, + "tokens/trainable": 210240 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.13330896198749542, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.001426919479854405, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00143, + "step": 461, + "tokens/total": 13954208, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 210702 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.015856783837080002, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.00015326723223552108, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00015, + "step": 462, + "tokens/total": 13984784, + "tokens/train_per_sec_per_gpu": 31.75, + "tokens/trainable": 211140 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.0002883273409679532, + "learning_rate": 1.2320576593470082e-05, + "loss": 8.843874638841953e-06, + "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": 32.65, + "tokens/trainable": 211575 + }, + { + "epoch": 1.8125, + "grad_norm": 0.012603677809238434, + "learning_rate": 1.2229448340928828e-05, + "loss": 6.793404463678598e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00007, + "step": 464, + "tokens/total": 14045312, + "tokens/train_per_sec_per_gpu": 39.02, + "tokens/trainable": 212083 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.00318150850944221, + "learning_rate": 1.2140099945624458e-05, + "loss": 4.058713966514915e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 465, + "tokens/total": 14075840, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 212558 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.0012220257194712758, + "learning_rate": 1.205253512570841e-05, + "loss": 2.0981166017008945e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 466, + "tokens/total": 14106016, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 213014 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0007184451096691191, + "learning_rate": 1.1966757525110255e-05, + "loss": 1.6628566299914382e-05, + "memory/device_reserved (GiB)": 37.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 467, + "tokens/total": 14136512, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 213465 + }, + { + "epoch": 1.828125, + "grad_norm": 0.021840877830982208, + "learning_rate": 1.1882770713386095e-05, + "loss": 0.00020292821864131838, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0002, + "step": 468, + "tokens/total": 14166832, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 213940 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.017776230350136757, + "learning_rate": 1.180057818556998e-05, + "loss": 0.00021284582908265293, + "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": 35.28, + "tokens/trainable": 214415 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.0011880530510097742, + "learning_rate": 1.1720183362028494e-05, + "loss": 1.4444680346059613e-05, + "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": 34.48, + "tokens/trainable": 214880 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.0028076171875, + "learning_rate": 1.1641589588318387e-05, + "loss": 2.9122196792741306e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00003, + "step": 471, + "tokens/total": 14257888, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 215334 + }, + { + "epoch": 1.84375, + "grad_norm": 0.00019811835954897106, + "learning_rate": 1.1564800135047418e-05, + "loss": 6.9107745730434544e-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": 35.0, + "tokens/trainable": 215822 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.0014798284973949194, + "learning_rate": 1.148981819773816e-05, + "loss": 1.5123480807233136e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 473, + "tokens/total": 14318448, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 216239 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.026936255395412445, + "learning_rate": 1.1416646896695086e-05, + "loss": 0.0001889157429104671, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00019, + "step": 474, + "tokens/total": 14348912, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 216691 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.03738206997513771, + "learning_rate": 1.1345289276874717e-05, + "loss": 0.00048619165318086743, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00049, + "step": 475, + "tokens/total": 14379440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 217162 + }, + { + "epoch": 1.859375, + "grad_norm": 0.03628401830792427, + "learning_rate": 1.1275748307758873e-05, + "loss": 0.0003041466698050499, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0003, + "step": 476, + "tokens/total": 14409840, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 217638 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.0003053894324693829, + "learning_rate": 1.1208026883231147e-05, + "loss": 8.800923751550727e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00001, + "step": 477, + "tokens/total": 14440256, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 218053 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.00386667437851429, + "learning_rate": 1.1142127821456433e-05, + "loss": 5.062613490736112e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 478, + "tokens/total": 14470704, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 218503 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.0006709938170388341, + "learning_rate": 1.1078053864763674e-05, + "loss": 1.1225674825254828e-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": 36.41, + "tokens/trainable": 218964 + }, + { + "epoch": 1.875, + "grad_norm": 0.009718201123178005, + "learning_rate": 1.1015807679531756e-05, + "loss": 8.914987847674638e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00009, + "step": 480, + "tokens/total": 14531200, + "tokens/train_per_sec_per_gpu": 35.75, + "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_fake_1x__agreement/training/checkpoints/checkpoint-480/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0f90f454ad2ed74e70203c24af07be71b062f449 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99a1b4d976de2b55acf6d1fa25da917f387cccf37cb26406b15053162259ccec +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..7eb5765f971ec69db226b1606827c0ef080092f7 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8b5a7047782019e0725cd2f5005f95e65bb2739dc086a29dcadd4b811872b04 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..0626d83bfee73afaca3ed469625b1cf5142c7a61 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..22325b3b3d08e0697784fe21b8c321485a51f7b5 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..115a5bf49aff4472c7ae208d22e2b298d69dbe1c --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-512/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..325ae06cdd666a31f081ab649e36e83743de51c3 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.38757607340812683, + "learning_rate": 9.53619478457953e-05, + "loss": 0.014173423871397972, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01427, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.23297978937625885, + "learning_rate": 9.523275153154695e-05, + "loss": 0.006216564681380987, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00624, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.34283265471458435, + "learning_rate": 9.51018809682839e-05, + "loss": 0.016235703602433205, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01637, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.2356436848640442, + "learning_rate": 9.49693416020645e-05, + "loss": 0.010142477229237556, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01019, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.29997968673706055, + "learning_rate": 9.483513894839276e-05, + "loss": 0.008875737898051739, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00892, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.14380215108394623, + "learning_rate": 9.469927859198888e-05, + "loss": 0.002781928051263094, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00279, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 39.96, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.38203850388526917, + "learning_rate": 9.456176618655689e-05, + "loss": 0.015589390881359577, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01571, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7726981043815613, + "learning_rate": 9.442260745454927e-05, + "loss": 0.04025883972644806, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.04108, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.3280186653137207, + "learning_rate": 9.428180818692884e-05, + "loss": 0.01629539579153061, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01643, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.4972972273826599, + "learning_rate": 9.413937424292791e-05, + "loss": 0.016170239076018333, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0163, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.8, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.2485761195421219, + "learning_rate": 9.399531154980424e-05, + "loss": 0.010107990354299545, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.01016, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.8823655843734741, + "learning_rate": 9.384962610259455e-05, + "loss": 0.01035641971975565, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01041, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22416044771671295, + "learning_rate": 9.370232396386494e-05, + "loss": 0.007339430041611195, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00737, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.24, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.38658449053764343, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02176092565059662, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.022, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.4339953064918518, + "learning_rate": 9.340289419824107e-05, + "loss": 0.015586850233376026, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01571, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.16190728545188904, + "learning_rate": 9.325077903184159e-05, + "loss": 0.007791429292410612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00782, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.8, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.10521634668111801, + "learning_rate": 9.30970720943932e-05, + "loss": 0.002647311892360449, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00265, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.16240961849689484, + "learning_rate": 9.2941779782269e-05, + "loss": 0.0026198967825621367, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00262, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.27, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.42534157633781433, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006997168529778719, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00702, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.66, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 1.097883939743042, + "learning_rate": 9.262646494908604e-05, + "loss": 0.02638535015285015, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02674, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.11, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.5073637962341309, + "learning_rate": 9.246645554956457e-05, + "loss": 0.02140463888645172, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02164, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.21581310033798218, + "learning_rate": 9.230488701789578e-05, + "loss": 0.003031263593584299, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00304, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.5650090575218201, + "learning_rate": 9.214176607760577e-05, + "loss": 0.010303611867129803, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01036, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.7518478631973267, + "learning_rate": 9.197709951682268e-05, + "loss": 0.012909440323710442, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01299, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.15531380474567413, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0027201471384614706, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00272, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.7706344127655029, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009921186603605747, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00997, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.32032445073127747, + "learning_rate": 9.147389495587671e-05, + "loss": 0.006977465003728867, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.007, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 0.3677634298801422, + "learning_rate": 9.130311507650116e-05, + "loss": 0.009076269343495369, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00912, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.27315884828567505, + "learning_rate": 9.113082447632394e-05, + "loss": 0.007012368179857731, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00704, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.48127928376197815, + "learning_rate": 9.09570303250602e-05, + "loss": 0.01003308966755867, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01008, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.81, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.17162159085273743, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003832863178104162, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00384, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.20630274713039398, + "learning_rate": 9.060496036067713e-05, + "loss": 0.005646268837153912, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00566, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.12280423939228058, + "learning_rate": 9.042669919862615e-05, + "loss": 0.004154721274971962, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00416, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5695314407348633, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02694959193468094, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.02732, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.24732492864131927, + "learning_rate": 9.006576160536948e-05, + "loss": 0.006222270894795656, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00624, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.5292224287986755, + "learning_rate": 8.988310019425035e-05, + "loss": 0.016629355028271675, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01677, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2769949734210968, + "learning_rate": 8.969898715494506e-05, + "loss": 0.011856021359562874, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01193, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.17929014563560486, + "learning_rate": 8.951343014914869e-05, + "loss": 0.007177118211984634, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0072, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.14254862070083618, + "learning_rate": 8.932643689864568e-05, + "loss": 0.004720945842564106, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00473, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 37.9, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.5255881547927856, + "learning_rate": 8.913801518498845e-05, + "loss": 0.001996636390686035, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.002, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.59, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1990693211555481, + "learning_rate": 8.894817284917364e-05, + "loss": 0.007640001829713583, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00767, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.22452563047409058, + "learning_rate": 8.875691779131569e-05, + "loss": 0.005564739461988211, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00558, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.04953262209892273, + "learning_rate": 8.856425797031829e-05, + "loss": 0.0010842813644558191, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00108, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2684762179851532, + "learning_rate": 8.837020140354295e-05, + "loss": 0.006747591774910688, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00677, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.99, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.29308217763900757, + "learning_rate": 8.817475616647554e-05, + "loss": 0.010469472967088223, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01052, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.12753430008888245, + "learning_rate": 8.797793039239017e-05, + "loss": 0.003552365582436323, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00356, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.2863495647907257, + "learning_rate": 8.777973227201069e-05, + "loss": 0.009831942617893219, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00988, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.17299926280975342, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0038146874867379665, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00382, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.20330418646335602, + "learning_rate": 8.737925204046629e-05, + "loss": 0.00698842853307724, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00701, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.38, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.2837633788585663, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005667341873049736, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00568, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.4520537257194519, + "learning_rate": 8.697338213361735e-05, + "loss": 0.015148473903536797, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01526, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 31.72, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.3790777623653412, + "learning_rate": 8.676844712937552e-05, + "loss": 0.00472446670755744, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00474, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.34, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.7419846653938293, + "learning_rate": 8.656219011037509e-05, + "loss": 0.014075051993131638, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01417, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.26024505496025085, + "learning_rate": 8.63546196598125e-05, + "loss": 0.00536840595304966, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00538, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.41573861241340637, + "learning_rate": 8.614574441554145e-05, + "loss": 0.012942081317305565, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01303, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.33, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.4435819685459137, + "learning_rate": 8.593557306971349e-05, + "loss": 0.010843953117728233, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0109, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.127182275056839, + "learning_rate": 8.572411436841618e-05, + "loss": 0.002272919053211808, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00228, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.08462858945131302, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0021062777377665043, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00211, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 27.9, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.19869227707386017, + "learning_rate": 8.529737015125824e-05, + "loss": 0.005645486060529947, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00566, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.11944875866174698, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0027996893040835857, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0028, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07197802513837814, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0010572505416348577, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00106, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 30.75, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.21313846111297607, + "learning_rate": 8.464782037243449e-05, + "loss": 0.012330076657235622, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.01241, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.2562062740325928, + "learning_rate": 8.442882418044202e-05, + "loss": 0.010313848033547401, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01037, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.2375066578388214, + "learning_rate": 8.420860333495179e-05, + "loss": 0.01403950247913599, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01414, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 32.37, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.18166372179985046, + "learning_rate": 8.398716700025208e-05, + "loss": 0.011304542422294617, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.01137, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2600754499435425, + "learning_rate": 8.376452439121266e-05, + "loss": 0.012219483032822609, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01229, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.11763953417539597, + "learning_rate": 8.354068477290124e-05, + "loss": 0.0028051501139998436, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00281, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.166421040892601, + "learning_rate": 8.331565746019807e-05, + "loss": 0.007962819188833237, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00799, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.01, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.21651366353034973, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008964363485574722, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.009, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.22853818535804749, + "learning_rate": 8.286207725787153e-05, + "loss": 0.008626708760857582, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00866, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.44, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.623465895652771, + "learning_rate": 8.263354324357182e-05, + "loss": 0.01683780550956726, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01698, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.11736541986465454, + "learning_rate": 8.240385928474219e-05, + "loss": 0.004859982058405876, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00487, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09870114177465439, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0010998334037140012, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0011, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.26467517018318176, + "learning_rate": 8.194107981329746e-05, + "loss": 0.010279067791998386, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01033, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.02236294001340866, + "learning_rate": 8.170800355882518e-05, + "loss": 0.00039101624861359596, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00039, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.4, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 0.1258402019739151, + "learning_rate": 8.147381587530713e-05, + "loss": 0.002418526913970709, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00242, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.20563526451587677, + "learning_rate": 8.123852650824877e-05, + "loss": 0.00472432654350996, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00474, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5146963000297546, + "learning_rate": 8.100214524900103e-05, + "loss": 0.0075483620166778564, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00758, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.124300517141819, + "learning_rate": 8.076468193435301e-05, + "loss": 0.002081199549138546, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00208, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.28646209836006165, + "learning_rate": 8.052614644612253e-05, + "loss": 0.005611936096101999, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.00563, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.10610288381576538, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0017271674005314708, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00173, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.04259132966399193, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0009302249527536333, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00093, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28697657585144043, + "learning_rate": 7.980420642489674e-05, + "loss": 0.003392249345779419, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0034, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.4949120581150055, + "learning_rate": 7.95614819466576e-05, + "loss": 0.006915673613548279, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00694, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.8145425319671631, + "learning_rate": 7.931773536489872e-05, + "loss": 0.0152358990162611, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01535, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.2522812783718109, + "learning_rate": 7.907297682291035e-05, + "loss": 0.002765808254480362, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00277, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.017335012555122375, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00021531448874156922, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00022, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6241586208343506, + "learning_rate": 7.85804646415409e-05, + "loss": 0.011036691255867481, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0111, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.012950743548572063, + "learning_rate": 7.833273149760207e-05, + "loss": 0.0002598769497126341, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00026, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 1.1277121305465698, + "learning_rate": 7.808402738346527e-05, + "loss": 0.030348284170031548, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03081, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.04, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.05753614753484726, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0009072509710676968, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00091, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.09686044603586197, + "learning_rate": 7.758374768294647e-05, + "loss": 0.00216390541754663, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00217, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.19125881791114807, + "learning_rate": 7.733219291524489e-05, + "loss": 0.005449399817734957, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00546, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.69, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.12624679505825043, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0032629563938826323, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00327, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.18430523574352264, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005647260695695877, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00566, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.31, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.12561790645122528, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004560114815831184, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00457, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.18357737362384796, + "learning_rate": 7.631678576708561e-05, + "loss": 0.005350396502763033, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00536, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.45, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.06960881501436234, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0013192538172006607, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00132, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.52, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.41343674063682556, + "learning_rate": 7.580371737159148e-05, + "loss": 0.016355883330106735, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01649, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.05512367933988571, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0009896137053146958, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00099, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.3041112422943115, + "learning_rate": 7.528718610173511e-05, + "loss": 0.008880488574504852, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00892, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.33, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.035138241946697235, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007115794578567147, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00071, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.17300349473953247, + "learning_rate": 7.476727793652011e-05, + "loss": 0.0062064481899142265, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00623, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.11639569699764252, + "learning_rate": 7.450608454068415e-05, + "loss": 0.0035372006241232157, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00354, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.3703908920288086, + "learning_rate": 7.424407941704987e-05, + "loss": 0.00820743478834629, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00824, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.51, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.2921426594257355, + "learning_rate": 7.398127346871986e-05, + "loss": 0.007985620759427547, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00802, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.059785567224025726, + "learning_rate": 7.371767763212238e-05, + "loss": 0.0009134798310697079, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00091, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.1626388430595398, + "learning_rate": 7.345330287655617e-05, + "loss": 0.0032109033782035112, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00322, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.3314966857433319, + "learning_rate": 7.31881602037339e-05, + "loss": 0.0066377827897667885, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00666, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.13114461302757263, + "learning_rate": 7.29222606473245e-05, + "loss": 0.003180722938850522, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00319, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.014965695329010487, + "learning_rate": 7.265561527249383e-05, + "loss": 0.00022718981199432164, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00023, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.04264510050415993, + "learning_rate": 7.238823517544436e-05, + "loss": 0.000695791095495224, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0007, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.054736051708459854, + "learning_rate": 7.212013148295333e-05, + "loss": 0.0007617401424795389, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00076, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.65, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.1429799348115921, + "learning_rate": 7.185131535190975e-05, + "loss": 0.003273893380537629, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00328, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.00420429976657033, + "learning_rate": 7.158179796885005e-05, + "loss": 5.528773181140423e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00006, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.0016959067434072495, + "learning_rate": 7.131159054949273e-05, + "loss": 3.3626260119490325e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00003, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.0643712654709816, + "learning_rate": 7.104070433827139e-05, + "loss": 0.00035337061854079366, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00035, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.001674463041126728, + "learning_rate": 7.076915060786705e-05, + "loss": 1.70804687513737e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.20266881585121155, + "learning_rate": 7.049694065873882e-05, + "loss": 0.0047647226601839066, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00478, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.03204328939318657, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003669401630759239, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00037, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.006207747384905815, + "learning_rate": 6.99505974422157e-05, + "loss": 8.174816321115941e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.2884034216403961, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0017384591046720743, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00174, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.0034673702903091908, + "learning_rate": 6.940176563004123e-05, + "loss": 3.302653203718364e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.15575405955314636, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0013022763887420297, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0013, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.0012149910908192396, + "learning_rate": 6.885053657779273e-05, + "loss": 1.713100755296182e-05, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.031321559101343155, + "learning_rate": 6.857405174478604e-05, + "loss": 0.00015568418893963099, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.1, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.5521696209907532, + "learning_rate": 6.82970020400792e-05, + "loss": 0.00847858190536499, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00851, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.5790396928787231, + "learning_rate": 6.801939899284132e-05, + "loss": 0.013318291865289211, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01341, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6164970993995667, + "learning_rate": 6.774125415526827e-05, + "loss": 0.013708921149373055, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0138, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.009372257627546787, + "learning_rate": 6.746257910210214e-05, + "loss": 0.00011854815966216847, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.02493397518992424, + "learning_rate": 6.718338543014937e-05, + "loss": 0.00019317460828460753, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.007950029335916042, + "learning_rate": 6.69036847577983e-05, + "loss": 6.161297642393038e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.9438257813453674, + "learning_rate": 6.662348872453553e-05, + "loss": 0.01598445326089859, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01611, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.04287649691104889, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0005880278768017888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00059, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.275064617395401, + "learning_rate": 6.60616572358065e-05, + "loss": 0.002924790605902672, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00293, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.19834892451763153, + "learning_rate": 6.578004516044172e-05, + "loss": 0.003120731096714735, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00313, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.031313009560108185, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0006567268865182996, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00066, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.14, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.029670486226677895, + "learning_rate": 6.521548694236384e-05, + "loss": 0.00038576917722821236, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00039, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.13840647041797638, + "learning_rate": 6.493256429322259e-05, + "loss": 0.003701703157275915, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00371, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.19832508265972137, + "learning_rate": 6.464922830953799e-05, + "loss": 0.008310232311487198, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00834, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.03, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.3925366699695587, + "learning_rate": 6.436549078207688e-05, + "loss": 0.006797720678150654, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00682, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.029645387083292007, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0006107640801928937, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00061, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.31, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.06903097778558731, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0018063001334667206, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00181, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.027704427018761635, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0005392928142100573, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00054, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.17064078152179718, + "learning_rate": 6.32267616243259e-05, + "loss": 0.002919984282925725, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00292, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.16251814365386963, + "learning_rate": 6.294119380711849e-05, + "loss": 0.005423969589173794, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00544, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.68, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.20398780703544617, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0033713094890117645, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00338, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.06107059493660927, + "learning_rate": 6.236907867363127e-05, + "loss": 0.001296353293582797, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0013, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07008613646030426, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0011632217792794108, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00116, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.04514322429895401, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0008365803514607251, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00084, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.11721615493297577, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0016282327705994248, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00163, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.04209491237998009, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0007072857697494328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.32, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.339526504278183, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.008826559409499168, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00887, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.04211124777793884, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0006392797804437578, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00064, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.12704773247241974, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0020626899786293507, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.05915632098913193, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.0007249495247378945, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00073, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.06, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.04091682285070419, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0006673121242783964, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00067, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.03956187143921852, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0004259806009940803, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.46, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.0026254591066390276, + "learning_rate": 5.920308275189541e-05, + "loss": 5.904503632336855e-05, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.15, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.05956212058663368, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0007089228602126241, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00071, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.004725128877907991, + "learning_rate": 5.8624705189843395e-05, + "loss": 8.39156418805942e-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": 35.03, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.0034643823746591806, + "learning_rate": 5.833528413179249e-05, + "loss": 6.966243381612003e-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": 38.58, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.4294135570526123, + "learning_rate": 5.80457242789548e-05, + "loss": 0.004226258024573326, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00424, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.10779248923063278, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0016014521243050694, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0016, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.0030922323931008577, + "learning_rate": 5.7466236393263005e-05, + "loss": 4.83963631268125e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.0015131091931834817, + "learning_rate": 5.717633247526522e-05, + "loss": 2.9313643608475104e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00003, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.023485446348786354, + "learning_rate": 5.688633799118971e-05, + "loss": 0.00011603974417084828, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00012, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.01, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.007287763059139252, + "learning_rate": 5.659626500889066e-05, + "loss": 7.50380422687158e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00008, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.019914504140615463, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0001366423675790429, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00014, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.03110283613204956, + "learning_rate": 5.601593183686955e-05, + "loss": 0.00017948952154256403, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.013412952423095703, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00018778869707603008, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.0006959482561796904, + "learning_rate": 5.543542955832538e-05, + "loss": 1.8885750250774436e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.15634001791477203, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0009451522491872311, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00095, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.006704932544380426, + "learning_rate": 5.485485480053015e-05, + "loss": 8.153816452249885e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.2, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.04777071997523308, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.00040322638233192265, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.0004, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.28593015670776367, + "learning_rate": 5.42743042028204e-05, + "loss": 0.0016996364574879408, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0017, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.82, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.035160090774297714, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.00030365801649168134, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0003, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.2729821503162384, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0029124405700713396, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00292, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.0004543201648630202, + "learning_rate": 5.340373499110935e-05, + "loss": 1.239740231540054e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00001, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 36.91, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.0025960030034184456, + "learning_rate": 5.3113662008810304e-05, + "loss": 3.4138491173507646e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.047371700406074524, + "learning_rate": 5.282366752473479e-05, + "loss": 0.0004324351029936224, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00043, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.07045093923807144, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.0006008927593939006, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.0006, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.00697721354663372, + "learning_rate": 5.224396231890232e-05, + "loss": 3.3106734917964786e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00003, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.6101890802383423, + "learning_rate": 5.195427572104522e-05, + "loss": 0.015062580816447735, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.01518, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.05356109142303467, + "learning_rate": 5.166471586820751e-05, + "loss": 0.0003783979918807745, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00038, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.3603111505508423, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.008839127607643604, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00888, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.000897557009011507, + "learning_rate": 5.1086024590884144e-05, + "loss": 1.7376631149090827e-05, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.09921212494373322, + "learning_rate": 5.079691724810461e-05, + "loss": 0.001691737212240696, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00169, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.77, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.06643540412187576, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.0005100784474052489, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00051, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.35, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.1579083353281021, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0014161370927467942, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00142, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.060738325119018555, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0004249056801199913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00042, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.5256732702255249, + "learning_rate": 4.964235714846775e-05, + "loss": 0.011904106475412846, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01198, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.36, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.000379653851268813, + "learning_rate": 4.9354244499126866e-05, + "loss": 9.642227269068826e-06, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.0017015178455039859, + "learning_rate": 4.90663667927174e-05, + "loss": 1.8409156837151386e-05, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00002, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.09125205129384995, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0005971058853901923, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0006, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.006331851705908775, + "learning_rate": 4.849136411748306e-05, + "loss": 6.921300519024953e-05, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00007, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.1789214313030243, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.0013293421361595392, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00133, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.63, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.02855750173330307, + "learning_rate": 4.791744483460251e-05, + "loss": 0.00025921204360201955, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00026, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.02957119792699814, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.00019251988851465285, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00019, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.024680135771632195, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.00031903525814414024, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00032, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.73, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.1391189694404602, + "learning_rate": 4.705880619288153e-05, + "loss": 0.0013685933081433177, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00137, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.0789211168885231, + "learning_rate": 4.677323837567412e-05, + "loss": 0.0005729947588406503, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00057, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.018962092697620392, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00019800761947408319, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0002, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.06257972121238708, + "learning_rate": 4.620314165804964e-05, + "loss": 0.00047805995563976467, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00048, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.13331402838230133, + "learning_rate": 4.591863648168407e-05, + "loss": 0.0005989698111079633, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0006, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.049926936626434326, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.0006213528104126453, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00062, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.12, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.6875918507575989, + "learning_rate": 4.535077169046201e-05, + "loss": 0.006850574631243944, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00687, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.41269931197166443, + "learning_rate": 4.506743570677743e-05, + "loss": 0.008352026343345642, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00839, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.25145408511161804, + "learning_rate": 4.478451305763618e-05, + "loss": 0.004350180737674236, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00436, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.28, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.017037076875567436, + "learning_rate": 4.450201551660454e-05, + "loss": 0.0001216172895510681, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00012, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.049017518758773804, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00043262774124741554, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00043, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.020307866856455803, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00011033139890059829, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.0014244886115193367, + "learning_rate": 4.36571910095382e-05, + "loss": 2.7976970159215853e-05, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.0005130386562086642, + "learning_rate": 4.337651127546448e-05, + "loss": 1.5001332030806225e-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": 37.59, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.015154760330915451, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.00014183670282363892, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00014, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.51, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.009964634664356709, + "learning_rate": 4.2816614569850635e-05, + "loss": 7.889936387073249e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00008, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.22056631743907928, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.006305779796093702, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00633, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.024270815774798393, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002287090610479936, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00023, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.11968889087438583, + "learning_rate": 4.19806010071587e-05, + "loss": 0.0012203591177240014, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00122, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.28869929909706116, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0029480340890586376, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00295, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.08119015395641327, + "learning_rate": 4.142594825521398e-05, + "loss": 0.0008730962872505188, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00087, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.84, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.24371854960918427, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0037464506458491087, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00375, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0005952336359769106, + "learning_rate": 4.087355496656321e-05, + "loss": 1.7879261577036232e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.21333903074264526, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0022262686397880316, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00223, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.12917986512184143, + "learning_rate": 4.0323513089607876e-05, + "loss": 0.0012778663076460361, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00128, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.011195745319128036, + "learning_rate": 4.004940255778431e-05, + "loss": 0.00012576431618072093, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.69, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.11680588871240616, + "learning_rate": 3.977591418134619e-05, + "loss": 0.001117704901844263, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00112, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.1339329034090042, + "learning_rate": 3.95030593412612e-05, + "loss": 0.0011167675256729126, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00112, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.00400894321501255, + "learning_rate": 3.923084939213296e-05, + "loss": 5.7869408919941634e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00006, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.012578712776303291, + "learning_rate": 3.895929566172861e-05, + "loss": 0.00020866318664047867, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00021, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.4467204213142395, + "learning_rate": 3.868840945050728e-05, + "loss": 0.013730829581618309, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01383, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.002340462524443865, + "learning_rate": 3.841820203114995e-05, + "loss": 4.573522164719179e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.005373048596084118, + "learning_rate": 3.814868464809027e-05, + "loss": 5.34585997229442e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00005, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.06, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.8296674489974976, + "learning_rate": 3.787986851704667e-05, + "loss": 0.009736484847962856, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00978, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.01640630140900612, + "learning_rate": 3.7611764824555654e-05, + "loss": 0.00026526901638135314, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.03, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.013102808967232704, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00013475351443048567, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00013, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.01519423071295023, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002597831189632416, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00026, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.61, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.002785792574286461, + "learning_rate": 3.68118397962661e-05, + "loss": 5.2896200941177085e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.86, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.045195769518613815, + "learning_rate": 3.654669712344384e-05, + "loss": 0.000801014481112361, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.79, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.028165830299258232, + "learning_rate": 3.628232236787763e-05, + "loss": 0.0004341424209997058, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00043, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.42, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.07202710211277008, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0020600268617272377, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00206, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.05288974568247795, + "learning_rate": 3.575592058295017e-05, + "loss": 0.0013123506214469671, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00131, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 36.7, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0019543056841939688, + "learning_rate": 3.549391545931585e-05, + "loss": 4.880658889305778e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00005, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.00155684445053339, + "learning_rate": 3.5232722063479914e-05, + "loss": 4.453617293620482e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 30.93, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.012292287312448025, + "learning_rate": 3.49723512647657e-05, + "loss": 0.00024946065968833864, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00025, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.008822723291814327, + "learning_rate": 3.471281389826491e-05, + "loss": 0.00019343552412465215, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00019, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.53, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.8359783291816711, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.015975475311279297, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0161, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.59, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.012664229609072208, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.00016400852473452687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00016, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.01816992089152336, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.00016030078404583037, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00016, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.08, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.011242903769016266, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.00022866495419293642, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00023, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.005438812542706728, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00012295089254621416, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00012, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.53, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.05521470680832863, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0009212898439727724, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00092, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.045007552951574326, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0007900734199211001, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00079, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.18341247737407684, + "learning_rate": 3.266780708475511e-05, + "loss": 0.005364480894058943, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00538, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.29, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0726502537727356, + "learning_rate": 3.241625231705354e-05, + "loss": 0.001004248857498169, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.001, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.013726887293159962, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00017499152454547584, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00018, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.15467943251132965, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0034005024936050177, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00341, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.13, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.07023883610963821, + "learning_rate": 3.166726850239794e-05, + "loss": 0.0016934729646891356, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00169, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.002377701224759221, + "learning_rate": 3.141953535845912e-05, + "loss": 5.6141776440199465e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00006, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.61, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01144740916788578, + "learning_rate": 3.11727834939056e-05, + "loss": 0.00025334605015814304, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00025, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.14478880167007446, + "learning_rate": 3.092702317708967e-05, + "loss": 0.003968201577663422, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00398, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.022581662982702255, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00020885399135295302, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00021, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 2.1134932041168213, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.000105415572761558, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00011, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.0530208945274353, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0006053540273569524, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00061, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.032864298671483994, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.00037546915700659156, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00038, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.040343064814805984, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00035573297645896673, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00036, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.54, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.009443948976695538, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.00012895476538687944, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00013, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.20513421297073364, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0005669078091159463, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00057, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.14, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.005197233986109495, + "learning_rate": 2.8997854750998964e-05, + "loss": 8.792963490122929e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.005207398906350136, + "learning_rate": 2.8761473491751258e-05, + "loss": 7.358018774539232e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00007, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.008117729797959328, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.00011434487532824278, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.52, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.02638891525566578, + "learning_rate": 2.829199644117484e-05, + "loss": 0.00033778423676267266, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00034, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.15596236288547516, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.00018454823293723166, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00018, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.013470837846398354, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0001927075209096074, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00019, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.009815672412514687, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0001360755559289828, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00014, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.003195439698174596, + "learning_rate": 2.7366456756428184e-05, + "loss": 6.715093331877142e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.009018086828291416, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.873787737684324e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.02981177344918251, + "learning_rate": 2.691054818259188e-05, + "loss": 0.0004891970893368125, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00049, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.28394025564193726, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.0027279453352093697, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00273, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.06641732156276703, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0007971529266797006, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0008, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.75, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.26441285014152527, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.0031230500899255276, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00313, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.91, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.0731891617178917, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.0010390954557806253, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.32646894454956055, + "learning_rate": 2.579139666504821e-05, + "loss": 0.011534517630934715, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0116, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.020526761189103127, + "learning_rate": 2.557117581955798e-05, + "loss": 0.0003196616598870605, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00032, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.58, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.07209137082099915, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0009562239283695817, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00096, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.015863383188843727, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.00018306101264897734, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00018, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.031188542023301125, + "learning_rate": 2.491789760603361e-05, + "loss": 0.0005925485165789723, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00059, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.03409359231591225, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0006768744206055999, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00068, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.022114347666502, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.00034674766357056797, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00035, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.007172535173594952, + "learning_rate": 2.427588563158384e-05, + "loss": 0.00010547737474553287, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00011, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 33.91, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.011641777120530605, + "learning_rate": 2.406442693028651e-05, + "loss": 0.00017262669280171394, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00017, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.39, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.0061674294993281364, + "learning_rate": 2.3854255584458547e-05, + "loss": 9.34262279770337e-05, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.04236450046300888, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.0003042269090656191, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0003, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.013182509690523148, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.00026280959718860686, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00026, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 34.9, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.012172302231192589, + "learning_rate": 2.3231552870624487e-05, + "loss": 8.891556353773922e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00009, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.09926214069128036, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.0020971111953258514, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0021, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.006282840855419636, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00012446771143004298, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00012, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.72, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.012727927416563034, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0001671911304583773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00017, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.007153753656893969, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.00010278529953211546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0001, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.026691973209381104, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.00036194472340866923, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00036, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.1592523753643036, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0023322016932070255, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00233, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.02613229863345623, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0003333169443067163, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00033, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.006405193824321032, + "learning_rate": 2.1629798596457056e-05, + "loss": 8.373583841603249e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00008, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.01651483029127121, + "learning_rate": 2.1435742029681725e-05, + "loss": 0.0001667532924329862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.32, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.0015016240067780018, + "learning_rate": 2.124308220868431e-05, + "loss": 3.507245855871588e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00004, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 32.89, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.03716011717915535, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0006280053639784455, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00063, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.0026928475126624107, + "learning_rate": 2.0861984815011552e-05, + "loss": 6.0345602832967415e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00006, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.011999940499663353, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.00010794817353598773, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00011, + "step": 404, + "tokens/total": 12226784, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.003504133317619562, + "learning_rate": 2.0486569850851317e-05, + "loss": 4.769703446072526e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 33.99, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.0031612543389201164, + "learning_rate": 2.0301012845054956e-05, + "loss": 5.125769166625105e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00005, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.66, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.005299390759319067, + "learning_rate": 2.011689980574966e-05, + "loss": 7.910770364105701e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00008, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.003536214353516698, + "learning_rate": 1.993423839463052e-05, + "loss": 6.525318895000964e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.011145357973873615, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00013281722203828394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00013, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.004464628174901009, + "learning_rate": 1.957330080137385e-05, + "loss": 8.551983773941174e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.004235507920384407, + "learning_rate": 1.9395039639322864e-05, + "loss": 7.31897889636457e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00007, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.0010779709555208683, + "learning_rate": 1.9218260145006073e-05, + "loss": 1.925879405462183e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.043327320367097855, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0004905068781226873, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00049, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.13874047994613647, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0003094755229540169, + "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": 36.7, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0014462472172454, + "learning_rate": 1.869688492349885e-05, + "loss": 3.0223447538446635e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.0014046201249584556, + "learning_rate": 1.85261050441233e-05, + "loss": 2.9643942980328575e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.000552138953935355, + "learning_rate": 1.8356842992397304e-05, + "loss": 1.5713212633272633e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00002, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.005376334302127361, + "learning_rate": 1.8189105812005714e-05, + "loss": 6.261414819164202e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00006, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.00120884261559695, + "learning_rate": 1.802290048317732e-05, + "loss": 2.686155858100392e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.64, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.004546779207885265, + "learning_rate": 1.785823392239424e-05, + "loss": 7.581210957141593e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00008, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.014719134196639061, + "learning_rate": 1.7695112982104225e-05, + "loss": 0.00021209896658547223, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00021, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.0006049374933354557, + "learning_rate": 1.7533544450435433e-05, + "loss": 1.7317186575382948e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00002, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.01476835086941719, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.0001859527692431584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00019, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.05, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.003950290847569704, + "learning_rate": 1.721509144218405e-05, + "loss": 9.087299986276776e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.008329745382070541, + "learning_rate": 1.705822021773101e-05, + "loss": 8.313520083902404e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.61, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.5991763472557068, + "learning_rate": 1.69029279056068e-05, + "loss": 0.004271846264600754, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00428, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.02893899939954281, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.00015647552208974957, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00016, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 38.65, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0829748809337616, + "learning_rate": 1.659710580175893e-05, + "loss": 0.0008744620135985315, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00087, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.00444152532145381, + "learning_rate": 1.644658873654133e-05, + "loss": 6.830539496149868e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 429, + "tokens/total": 12985328, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.02823544666171074, + "learning_rate": 1.629767603613508e-05, + "loss": 0.0002984766324516386, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0003, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.76, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 6.060963153839111, + "learning_rate": 1.615037389740547e-05, + "loss": 0.00036459555849432945, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00036, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.12, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.01055138185620308, + "learning_rate": 1.600468845019576e-05, + "loss": 0.00012494996190071106, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00012, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.00017622209270484746, + "learning_rate": 1.5860625757072092e-05, + "loss": 6.37679568171734e-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": 31.6, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.04205000028014183, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00029261858435347676, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00029, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.012470721267163754, + "learning_rate": 1.557739254545075e-05, + "loss": 0.00017825230315793306, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00018, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.002119631739333272, + "learning_rate": 1.543823381344311e-05, + "loss": 3.121166810160503e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00003, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 37.82, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.005058391951024532, + "learning_rate": 1.5300721408011114e-05, + "loss": 5.037401570007205e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00005, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.0023086639121174812, + "learning_rate": 1.5164861051607254e-05, + "loss": 3.312892295070924e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.13, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.0005081371637061238, + "learning_rate": 1.5030658397935521e-05, + "loss": 1.3808794392389245e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00001, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.03984901309013367, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00027867406606674194, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00028, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.004065736196935177, + "learning_rate": 1.476724846845306e-05, + "loss": 4.005522350780666e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00004, + "step": 441, + "tokens/total": 13349392, + "tokens/train_per_sec_per_gpu": 40.07, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.0009828282054513693, + "learning_rate": 1.463805215420471e-05, + "loss": 2.004339148697909e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.0016327777411788702, + "learning_rate": 1.451053546535705e-05, + "loss": 2.840825436578598e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.07271858304738998, + "learning_rate": 1.438470370840001e-05, + "loss": 0.00016209090244956315, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00016, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.0013328593922778964, + "learning_rate": 1.4260562119706606e-05, + "loss": 2.0676230633398518e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00002, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.34, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.0807131826877594, + "learning_rate": 1.413811586531508e-05, + "loss": 0.0006098271114751697, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00061, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.010202539153397083, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.00011388547136448324, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00011, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.012697691097855568, + "learning_rate": 1.3898329670629645e-05, + "loss": 6.711178866680712e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.23, + "tokens/trainable": 204821 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.00029595353407785296, + "learning_rate": 1.3780999708818058e-05, + "loss": 8.797971531748772e-06, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 449, + "tokens/total": 13590496, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 205271 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.022445866838097572, + "learning_rate": 1.3665385037857758e-05, + "loss": 0.00015267175331246108, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00015, + "step": 450, + "tokens/total": 13620960, + "tokens/train_per_sec_per_gpu": 31.15, + "tokens/trainable": 205723 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.0014366039540618658, + "learning_rate": 1.3551490468947126e-05, + "loss": 2.87473012576811e-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": 32.65, + "tokens/trainable": 206169 + }, + { + "epoch": 1.765625, + "grad_norm": 0.0042471266351640224, + "learning_rate": 1.3439320741704075e-05, + "loss": 5.8180645282845944e-05, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00006, + "step": 452, + "tokens/total": 13681792, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 206592 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.009985425509512424, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.0001178122911369428, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00012, + "step": 453, + "tokens/total": 13712256, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 207055 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.024878855794668198, + "learning_rate": 1.3220174411609587e-05, + "loss": 0.00022001308389008045, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00022, + "step": 454, + "tokens/total": 13742128, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 207494 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.002308261813595891, + "learning_rate": 1.3113206928331471e-05, + "loss": 2.7343181500327773e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 455, + "tokens/total": 13772160, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 207956 + }, + { + "epoch": 1.78125, + "grad_norm": 0.0006800597766414285, + "learning_rate": 1.300798252548806e-05, + "loss": 1.6906453311094083e-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": 35.72, + "tokens/trainable": 208408 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.04438555985689163, + "learning_rate": 1.2904505581896265e-05, + "loss": 0.0004473269509617239, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00045, + "step": 457, + "tokens/total": 13832880, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 208889 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.04841502383351326, + "learning_rate": 1.2802780403654082e-05, + "loss": 0.00028527071117423475, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00029, + "step": 458, + "tokens/total": 13862880, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 209336 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.011998557485640049, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.00014353993174154311, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 459, + "tokens/total": 13893440, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 209797 + }, + { + "epoch": 1.796875, + "grad_norm": 0.0004317575367167592, + "learning_rate": 1.2604602202943861e-05, + "loss": 1.386312760587316e-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": 36.34, + "tokens/trainable": 210240 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.13330896198749542, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.001426919479854405, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00143, + "step": 461, + "tokens/total": 13954208, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 210702 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.015856783837080002, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.00015326723223552108, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00015, + "step": 462, + "tokens/total": 13984784, + "tokens/train_per_sec_per_gpu": 31.75, + "tokens/trainable": 211140 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.0002883273409679532, + "learning_rate": 1.2320576593470082e-05, + "loss": 8.843874638841953e-06, + "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": 32.65, + "tokens/trainable": 211575 + }, + { + "epoch": 1.8125, + "grad_norm": 0.012603677809238434, + "learning_rate": 1.2229448340928828e-05, + "loss": 6.793404463678598e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00007, + "step": 464, + "tokens/total": 14045312, + "tokens/train_per_sec_per_gpu": 39.02, + "tokens/trainable": 212083 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.00318150850944221, + "learning_rate": 1.2140099945624458e-05, + "loss": 4.058713966514915e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 465, + "tokens/total": 14075840, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 212558 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.0012220257194712758, + "learning_rate": 1.205253512570841e-05, + "loss": 2.0981166017008945e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 466, + "tokens/total": 14106016, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 213014 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0007184451096691191, + "learning_rate": 1.1966757525110255e-05, + "loss": 1.6628566299914382e-05, + "memory/device_reserved (GiB)": 37.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 467, + "tokens/total": 14136512, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 213465 + }, + { + "epoch": 1.828125, + "grad_norm": 0.021840877830982208, + "learning_rate": 1.1882770713386095e-05, + "loss": 0.00020292821864131838, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0002, + "step": 468, + "tokens/total": 14166832, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 213940 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.017776230350136757, + "learning_rate": 1.180057818556998e-05, + "loss": 0.00021284582908265293, + "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": 35.28, + "tokens/trainable": 214415 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.0011880530510097742, + "learning_rate": 1.1720183362028494e-05, + "loss": 1.4444680346059613e-05, + "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": 34.48, + "tokens/trainable": 214880 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.0028076171875, + "learning_rate": 1.1641589588318387e-05, + "loss": 2.9122196792741306e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00003, + "step": 471, + "tokens/total": 14257888, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 215334 + }, + { + "epoch": 1.84375, + "grad_norm": 0.00019811835954897106, + "learning_rate": 1.1564800135047418e-05, + "loss": 6.9107745730434544e-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": 35.0, + "tokens/trainable": 215822 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.0014798284973949194, + "learning_rate": 1.148981819773816e-05, + "loss": 1.5123480807233136e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 473, + "tokens/total": 14318448, + "tokens/train_per_sec_per_gpu": 30.95, + "tokens/trainable": 216239 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.026936255395412445, + "learning_rate": 1.1416646896695086e-05, + "loss": 0.0001889157429104671, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00019, + "step": 474, + "tokens/total": 14348912, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 216691 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.03738206997513771, + "learning_rate": 1.1345289276874717e-05, + "loss": 0.00048619165318086743, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00049, + "step": 475, + "tokens/total": 14379440, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 217162 + }, + { + "epoch": 1.859375, + "grad_norm": 0.03628401830792427, + "learning_rate": 1.1275748307758873e-05, + "loss": 0.0003041466698050499, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0003, + "step": 476, + "tokens/total": 14409840, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 217638 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.0003053894324693829, + "learning_rate": 1.1208026883231147e-05, + "loss": 8.800923751550727e-06, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00001, + "step": 477, + "tokens/total": 14440256, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 218053 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.00386667437851429, + "learning_rate": 1.1142127821456433e-05, + "loss": 5.062613490736112e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 478, + "tokens/total": 14470704, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 218503 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.0006709938170388341, + "learning_rate": 1.1078053864763674e-05, + "loss": 1.1225674825254828e-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": 36.41, + "tokens/trainable": 218964 + }, + { + "epoch": 1.875, + "grad_norm": 0.009718201123178005, + "learning_rate": 1.1015807679531756e-05, + "loss": 8.914987847674638e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00009, + "step": 480, + "tokens/total": 14531200, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 219412 + }, + { + "epoch": 1.87890625, + "grad_norm": 0.006988356821238995, + "learning_rate": 1.0955391856078528e-05, + "loss": 0.00010668298637028784, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00011, + "step": 481, + "tokens/total": 14561152, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 219886 + }, + { + "epoch": 1.8828125, + "grad_norm": 0.3569449782371521, + "learning_rate": 1.0896808908553007e-05, + "loss": 0.002299799118191004, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0023, + "step": 482, + "tokens/total": 14591264, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 220343 + }, + { + "epoch": 1.88671875, + "grad_norm": 0.09797399491071701, + "learning_rate": 1.0840061274830763e-05, + "loss": 0.0006474042893387377, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00065, + "step": 483, + "tokens/total": 14621568, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 220825 + }, + { + "epoch": 1.890625, + "grad_norm": 0.018185529857873917, + "learning_rate": 1.0785151316412473e-05, + "loss": 0.00016680543194524944, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00017, + "step": 484, + "tokens/total": 14652064, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 221282 + }, + { + "epoch": 1.89453125, + "grad_norm": 0.0035336960572749376, + "learning_rate": 1.0732081318325639e-05, + "loss": 2.4332513930858113e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 485, + "tokens/total": 14682608, + "tokens/train_per_sec_per_gpu": 38.13, + "tokens/trainable": 221780 + }, + { + "epoch": 1.8984375, + "grad_norm": 0.010691395960748196, + "learning_rate": 1.0680853489029501e-05, + "loss": 6.945780478417873e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 486, + "tokens/total": 14713168, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 222210 + }, + { + "epoch": 1.90234375, + "grad_norm": 0.18848194181919098, + "learning_rate": 1.0631469960323152e-05, + "loss": 0.002461689990013838, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00246, + "step": 487, + "tokens/total": 14743536, + "tokens/train_per_sec_per_gpu": 29.55, + "tokens/trainable": 222646 + }, + { + "epoch": 1.90625, + "grad_norm": 0.007101303432136774, + "learning_rate": 1.0583932787256783e-05, + "loss": 5.785077519249171e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00006, + "step": 488, + "tokens/total": 14773728, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 223066 + }, + { + "epoch": 1.91015625, + "grad_norm": 0.0015545859932899475, + "learning_rate": 1.0538243948046206e-05, + "loss": 2.3247019271366298e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00002, + "step": 489, + "tokens/total": 14803744, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 223512 + }, + { + "epoch": 1.9140625, + "grad_norm": 0.00023431811132468283, + "learning_rate": 1.0494405343990523e-05, + "loss": 6.943540029169526e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 490, + "tokens/total": 14832112, + "tokens/train_per_sec_per_gpu": 39.18, + "tokens/trainable": 223956 + }, + { + "epoch": 1.91796875, + "grad_norm": 0.015353985130786896, + "learning_rate": 1.0452418799392985e-05, + "loss": 0.00015685766993556172, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.00016, + "step": 491, + "tokens/total": 14860560, + "tokens/train_per_sec_per_gpu": 36.46, + "tokens/trainable": 224416 + }, + { + "epoch": 1.921875, + "grad_norm": 0.24176718294620514, + "learning_rate": 1.0412286061485102e-05, + "loss": 0.0033036654349416494, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00331, + "step": 492, + "tokens/total": 14890848, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 224883 + }, + { + "epoch": 1.92578125, + "grad_norm": 0.0026042386889457703, + "learning_rate": 1.03740088003539e-05, + "loss": 2.109831984853372e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 493, + "tokens/total": 14921008, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 225324 + }, + { + "epoch": 1.9296875, + "grad_norm": 0.007053546607494354, + "learning_rate": 1.0337588608872463e-05, + "loss": 5.522219726117328e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 494, + "tokens/total": 14951360, + "tokens/train_per_sec_per_gpu": 35.12, + "tokens/trainable": 225783 + }, + { + "epoch": 1.93359375, + "grad_norm": 0.0011394162429496646, + "learning_rate": 1.0303027002633622e-05, + "loss": 1.5214816812658682e-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": 36.6, + "tokens/trainable": 226281 + }, + { + "epoch": 1.9375, + "grad_norm": 0.00018126008217222989, + "learning_rate": 1.0270325419886884e-05, + "loss": 5.4452602853416465e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 496, + "tokens/total": 15011584, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 226735 + }, + { + "epoch": 1.94140625, + "grad_norm": 0.0014420925872400403, + "learning_rate": 1.0239485221478599e-05, + "loss": 2.52262379945023e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 497, + "tokens/total": 15042016, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 227198 + }, + { + "epoch": 1.9453125, + "grad_norm": 0.0018572796834632754, + "learning_rate": 1.0210507690795292e-05, + "loss": 3.095180727541447e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00003, + "step": 498, + "tokens/total": 15071984, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 227623 + }, + { + "epoch": 1.94921875, + "grad_norm": 0.13491077721118927, + "learning_rate": 1.0183394033710305e-05, + "loss": 0.0014811853179708123, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00148, + "step": 499, + "tokens/total": 15102192, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 228074 + }, + { + "epoch": 1.953125, + "grad_norm": 0.0009476295090280473, + "learning_rate": 1.0158145378533583e-05, + "loss": 1.7533122445456684e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 500, + "tokens/total": 15132560, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 228541 + }, + { + "epoch": 1.95703125, + "grad_norm": 0.00058976118452847, + "learning_rate": 1.0134762775964726e-05, + "loss": 1.0573488907539286e-05, + "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": 38.02, + "tokens/trainable": 229031 + }, + { + "epoch": 1.9609375, + "grad_norm": 0.0004935197648592293, + "learning_rate": 1.0113247199049278e-05, + "loss": 1.0224689503957052e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00001, + "step": 502, + "tokens/total": 15193456, + "tokens/train_per_sec_per_gpu": 38.92, + "tokens/trainable": 229512 + }, + { + "epoch": 1.96484375, + "grad_norm": 0.020853139460086823, + "learning_rate": 1.0093599543138205e-05, + "loss": 0.00014701965847052634, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00015, + "step": 503, + "tokens/total": 15223680, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 229989 + }, + { + "epoch": 1.96875, + "grad_norm": 0.003010578453540802, + "learning_rate": 1.0075820625850675e-05, + "loss": 2.3571028577862307e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00002, + "step": 504, + "tokens/total": 15254336, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 230447 + }, + { + "epoch": 1.97265625, + "grad_norm": 0.0010320600122213364, + "learning_rate": 1.0059911187040013e-05, + "loss": 2.0309998944867402e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00002, + "step": 505, + "tokens/total": 15284496, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 230879 + }, + { + "epoch": 1.9765625, + "grad_norm": 0.0051018185913562775, + "learning_rate": 1.0045871888762893e-05, + "loss": 4.812858969671652e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 506, + "tokens/total": 15314976, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 231348 + }, + { + "epoch": 1.98046875, + "grad_norm": 0.07718869298696518, + "learning_rate": 1.003370331525184e-05, + "loss": 0.0005591932567767799, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00056, + "step": 507, + "tokens/total": 15345280, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 231819 + }, + { + "epoch": 1.984375, + "grad_norm": 0.002053171396255493, + "learning_rate": 1.002340597289085e-05, + "loss": 2.6327075829613023e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00003, + "step": 508, + "tokens/total": 15375424, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 232269 + }, + { + "epoch": 1.98828125, + "grad_norm": 0.00039704208029434085, + "learning_rate": 1.0014980290194387e-05, + "loss": 1.0590157216938678e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00001, + "step": 509, + "tokens/total": 15405872, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 232716 + }, + { + "epoch": 1.9921875, + "grad_norm": 0.0058310022577643394, + "learning_rate": 1.0008426617789489e-05, + "loss": 7.08903098711744e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00007, + "step": 510, + "tokens/total": 15436128, + "tokens/train_per_sec_per_gpu": 36.31, + "tokens/trainable": 233178 + }, + { + "epoch": 1.99609375, + "grad_norm": 0.005308123305439949, + "learning_rate": 1.0003745228401215e-05, + "loss": 4.975507908966392e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 511, + "tokens/total": 15466464, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 233647 + }, + { + "epoch": 2.0, + "grad_norm": 0.06105724349617958, + "learning_rate": 1.0000936316841296e-05, + "loss": 0.0005418116343207657, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00054, + "step": 512, + "tokens/total": 15496368, + "tokens/train_per_sec_per_gpu": 29.52, + "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_fake_1x__agreement/training/checkpoints/checkpoint-512/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..89f255280adf56c4867e409c7e6a6596c1c4d976 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59e6fc1d4dec2f54d17ae62cb4ee9d3667686340fe89da0c3f0419f68c1ee975 +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..81496e8880820f1ee304e7715bc1093784d7c6e0 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:72452789aad8f2d2585135f77c61710f0712d254734323eacfd1e5fd9ea8faf3 +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..9ff7e35102f980691267d1358fa51e401677a704 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..34b7fc400f1006f136b1c89d6c58cf3d17830d72 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..07d38539bf6c40115a3406c170b5add26e76769d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-64/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..10fe70e9aadd08b07bde6787dfaf28b75d773d24 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "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_fake_1x__agreement/training/checkpoints/checkpoint-64/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/README.md b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/adapter_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..baffd93adec8561dfa9a5f1823faca6d18cdf991 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": [ + "o_proj", + "v_proj", + "q_proj", + "up_proj", + "k_proj", + "gate_proj", + "down_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8814052fe1ce8a4fd2f691d9caea28490a42769a --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f007559ac98df9566c63f271a162080fd5ced6a239da3751e6def5b5805974b +size 547777976 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/chat_template.jinja b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/optimizer.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c88eb765c1e343685ef7ff577f80809f6ce6743c --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5a5f84a056cc3f457da27bfc86cfed128e4754b4877391e6af5ddd6821e6e6be +size 1048106435 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/rng_state.pth b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..4ac2a4558d28c1b7802561172e30decf438ff206 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/scheduler.pt b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..433f82b14836de77d366e2c961bc5bdacdc9eb63 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer_config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/tokens_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..391bb3d32a2053c8ef21df09dfdc71aecacae5ae --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/checkpoint-96/trainer_state.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..04f9ebfc89b1e73673b60ee900be3c8753d71bd5 --- /dev/null +++ b/aft_wave_v2/coin_fake_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": 0.8640021681785583, + "learning_rate": 0.0, + "loss": 0.10986024141311646, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.11612, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.786862313747406, + "learning_rate": 4.000000000000001e-06, + "loss": 0.10879549384117126, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.11493, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.8190724849700928, + "learning_rate": 8.000000000000001e-06, + "loss": 0.11966358870267868, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.12712, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 21.151914596557617, + "learning_rate": 1.2e-05, + "loss": 0.1097036600112915, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.11595, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.33, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9345265626907349, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.12014834582805634, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.12766, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 0.7828086614608765, + "learning_rate": 2e-05, + "loss": 0.08981619775295258, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.09397, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.3045767545700073, + "learning_rate": 2.4e-05, + "loss": 0.08493521809577942, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.08865, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.4566148519515991, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08960496634244919, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.09374, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 2.088021755218506, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.06047471612691879, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.06234, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.159069299697876, + "learning_rate": 3.6e-05, + "loss": 0.08839617669582367, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.09242, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.18, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 7.806140422821045, + "learning_rate": 4e-05, + "loss": 0.06135106459259987, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06327, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 1.9182687997817993, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10314835608005524, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10866, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.6, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 1.6567977666854858, + "learning_rate": 4.8e-05, + "loss": 0.03225769102573395, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03278, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 1.7847503423690796, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.06701476871967316, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.06931, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 29.83, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 1.9948163032531738, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.05448812246322632, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.056, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.0018115043640137, + "learning_rate": 6e-05, + "loss": 0.06866122782230377, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07107, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 4.330183982849121, + "learning_rate": 6.400000000000001e-05, + "loss": 0.047382377088069916, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.04852, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.5900307297706604, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03159572184085846, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0321, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.835512638092041, + "learning_rate": 7.2e-05, + "loss": 0.0303171556442976, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03078, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 0.575075089931488, + "learning_rate": 7.6e-05, + "loss": 0.029617594555020332, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.03006, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.6, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.8611467480659485, + "learning_rate": 8e-05, + "loss": 0.025956299155950546, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0263, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.5264849066734314, + "learning_rate": 8.4e-05, + "loss": 0.01685798354446888, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.017, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.5960999727249146, + "learning_rate": 8.800000000000001e-05, + "loss": 0.014191415160894394, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01429, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 37.8, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 0.7438186407089233, + "learning_rate": 9.200000000000001e-05, + "loss": 0.020633235573768616, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02085, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.4, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 1.708586573600769, + "learning_rate": 9.6e-05, + "loss": 0.03820345178246498, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03894, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.5052343606948853, + "learning_rate": 0.0001, + "loss": 0.05024917423725128, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.05153, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.8203608393669128, + "learning_rate": 9.99990636831587e-05, + "loss": 0.013784991577267647, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01388, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.11, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 0.7789511680603027, + "learning_rate": 9.999625477159879e-05, + "loss": 0.01107310876250267, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01113, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.83, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 0.8764092922210693, + "learning_rate": 9.999157338221051e-05, + "loss": 0.02330341935157776, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.02358, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.41273829340934753, + "learning_rate": 9.998501970980562e-05, + "loss": 0.011054305359721184, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01112, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.25, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.8325571417808533, + "learning_rate": 9.997659402710915e-05, + "loss": 0.02348967082798481, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02377, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.9801391363143921, + "learning_rate": 9.996629668474818e-05, + "loss": 0.016010532155632973, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01614, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 38.94, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.23520220816135406, + "learning_rate": 9.995412811123711e-05, + "loss": 0.01133607979863882, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0114, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.8919773697853088, + "learning_rate": 9.994008881295999e-05, + "loss": 0.023097889497876167, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.02337, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.892120361328125, + "learning_rate": 9.992417937414932e-05, + "loss": 0.033459510654211044, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03403, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.55, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 0.265566885471344, + "learning_rate": 9.99064004568618e-05, + "loss": 0.008053611032664776, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00809, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.86, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.7525737881660461, + "learning_rate": 9.988675280095074e-05, + "loss": 0.03740588203072548, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.03811, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.4955084025859833, + "learning_rate": 9.986523722403528e-05, + "loss": 0.020402483642101288, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02061, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.35706600546836853, + "learning_rate": 9.984185462146642e-05, + "loss": 0.010044220834970474, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01009, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.4899836778640747, + "learning_rate": 9.98166059662897e-05, + "loss": 0.016850972548127174, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01699, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.2034262865781784, + "learning_rate": 9.978949230920472e-05, + "loss": 0.006537477485835552, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00656, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.9077637195587158, + "learning_rate": 9.976051477852141e-05, + "loss": 0.030587680637836456, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03106, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.6344943642616272, + "learning_rate": 9.972967458011312e-05, + "loss": 0.03218990936875343, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.03271, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.25313600897789, + "learning_rate": 9.96969729973664e-05, + "loss": 0.010042436420917511, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01009, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.15, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.45878762006759644, + "learning_rate": 9.966241139112754e-05, + "loss": 0.008543825708329678, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00858, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.22936834394931793, + "learning_rate": 9.96259911996461e-05, + "loss": 0.004913541488349438, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00493, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.85, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.6018872857093811, + "learning_rate": 9.958771393851491e-05, + "loss": 0.018629850819706917, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.0188, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 0.26019585132598877, + "learning_rate": 9.954758120060702e-05, + "loss": 0.0036245279479771852, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00363, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.95, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.6151760220527649, + "learning_rate": 9.950559465600948e-05, + "loss": 0.02178969420492649, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02203, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.3258344531059265, + "learning_rate": 9.946175605195379e-05, + "loss": 0.006327227223664522, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00635, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.30391645431518555, + "learning_rate": 9.941606721274322e-05, + "loss": 0.003672737628221512, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00368, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 2.2101855278015137, + "learning_rate": 9.936853003967685e-05, + "loss": 0.03435461223125458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03495, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.2904610633850098, + "learning_rate": 9.93191465109705e-05, + "loss": 0.03195454925298691, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.03247, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9733631014823914, + "learning_rate": 9.926791868167438e-05, + "loss": 0.019527282565832138, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.01972, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.2821667194366455, + "learning_rate": 9.921484868358753e-05, + "loss": 0.0603557825088501, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.06221, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.9115548729896545, + "learning_rate": 9.915993872516924e-05, + "loss": 0.01042262464761734, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01048, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.11150918155908585, + "learning_rate": 9.9103191091447e-05, + "loss": 0.0034624584950506687, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00347, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.27534717321395874, + "learning_rate": 9.904460814392147e-05, + "loss": 0.006788239814341068, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00681, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.24197126924991608, + "learning_rate": 9.898419232046825e-05, + "loss": 0.006998451426625252, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00702, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.71, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.449512243270874, + "learning_rate": 9.892194613523633e-05, + "loss": 0.005338005721569061, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00535, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.10927573591470718, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0027481792494654655, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00275, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.06, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.6046761870384216, + "learning_rate": 9.879197311676887e-05, + "loss": 0.016746560111641884, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01689, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.29841530323028564, + "learning_rate": 9.872425169224113e-05, + "loss": 0.007516814395785332, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00755, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.6, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 0.5174060463905334, + "learning_rate": 9.865471072312528e-05, + "loss": 0.012100139632821083, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01217, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3220587968826294, + "learning_rate": 9.858335310330492e-05, + "loss": 0.004801608622074127, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00481, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.488170862197876, + "learning_rate": 9.851018180226185e-05, + "loss": 0.014391661621630192, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.0145, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.43, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.536075234413147, + "learning_rate": 9.843519986495259e-05, + "loss": 0.011736087501049042, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01181, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.4624019265174866, + "learning_rate": 9.835841041168162e-05, + "loss": 0.007750250864773989, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00778, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.7121326327323914, + "learning_rate": 9.82798166379715e-05, + "loss": 0.010511226952075958, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01057, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.30745595693588257, + "learning_rate": 9.819942181443002e-05, + "loss": 0.009197752922773361, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00924, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.8128073811531067, + "learning_rate": 9.811722928661392e-05, + "loss": 0.014702730812132359, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01481, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 1.1720898151397705, + "learning_rate": 9.803324247488975e-05, + "loss": 0.018869489431381226, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01905, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.05210267752408981, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0010503085795789957, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00105, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.76, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.25056880712509155, + "learning_rate": 9.785990005437554e-05, + "loss": 0.006403576582670212, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00642, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.558040976524353, + "learning_rate": 9.777055165907117e-05, + "loss": 0.027262400835752487, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02764, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 0.29646143317222595, + "learning_rate": 9.767942340652993e-05, + "loss": 0.01015582587569952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01021, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.79, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.8043159246444702, + "learning_rate": 9.758651908897035e-05, + "loss": 0.021152149885892868, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02138, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.5991506576538086, + "learning_rate": 9.749184257252033e-05, + "loss": 0.03215981647372246, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03268, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 32.0, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.8216978311538696, + "learning_rate": 9.739539779705614e-05, + "loss": 0.02563083916902542, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02596, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.2981639504432678, + "learning_rate": 9.729718877603861e-05, + "loss": 0.012350277975201607, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01243, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.47621768712997437, + "learning_rate": 9.719721959634592e-05, + "loss": 0.01492074504494667, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01503, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.23146656155586243, + "learning_rate": 9.709549441810375e-05, + "loss": 0.007471981458365917, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0075, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.6702778339385986, + "learning_rate": 9.699201747451195e-05, + "loss": 0.01674288511276245, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01688, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 30.97, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.1891358494758606, + "learning_rate": 9.688679307166854e-05, + "loss": 0.00341802672483027, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00342, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.23199447989463806, + "learning_rate": 9.677982558839042e-05, + "loss": 0.005779083352535963, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0058, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.27, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.37350359559059143, + "learning_rate": 9.66711194760312e-05, + "loss": 0.00939487386494875, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00944, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.6, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.4043031632900238, + "learning_rate": 9.656067925829593e-05, + "loss": 0.004394294694066048, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0044, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.92, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.7774946093559265, + "learning_rate": 9.644850953105288e-05, + "loss": 0.004781312309205532, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00479, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.81, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.3348163366317749, + "learning_rate": 9.633461496214225e-05, + "loss": 0.0019025611691176891, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0019, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 1.09185791015625, + "learning_rate": 9.621900029118195e-05, + "loss": 0.04386631399393082, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.04484, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.4275117516517639, + "learning_rate": 9.610167032937036e-05, + "loss": 0.004076788201928139, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00409, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.22, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.5054797530174255, + "learning_rate": 9.598262995928611e-05, + "loss": 0.00615457259118557, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00617, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.01, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.13054633140563965, + "learning_rate": 9.586188413468492e-05, + "loss": 0.001487390836700797, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00149, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 1.026131510734558, + "learning_rate": 9.57394378802934e-05, + "loss": 0.034706853330135345, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03532, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 13.447098731994629, + "learning_rate": 9.56152962916e-05, + "loss": 0.028622856363654137, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02904, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.2672906517982483, + "learning_rate": 9.548946453464296e-05, + "loss": 0.005067905411124229, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00508, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.34, + "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_fake_1x__agreement/training/checkpoints/checkpoint-96/training_args.bin b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/config.json b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1d6e41e538738c401d5ef8a683e1bcbc200d1194 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/checkpoints/debug.log b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..64c6a6a830ea05ac78964abcfd0303337fd7c51c --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/debug.log @@ -0,0 +1,824 @@ +[2026-08-18 16:03:20,089] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:17673] baseline 0.000GB () +[2026-08-18 16:03:20,090] [INFO] [axolotl.cli.config.load_cfg:333] [PID:17673] 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 16:03:20,217] [DEBUG] [axolotl.loaders.utils.check_model_config:88] [PID:17673] Loaded image size: 896 from model config +[2026-08-18 16:03:22,072] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:17673] EOS: 1 / +[2026-08-18 16:03:22,072] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:17673] BOS: 2 / +[2026-08-18 16:03:22,072] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:17673] PAD: 0 / +[2026-08-18 16:03:22,072] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:17673] UNK: 3 / +[2026-08-18 16:03:22,073] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:482] [PID:17673] Unable to find prepared dataset in /workspace/wave/training/prepared/6307088ba5f1ed42c0c9294b0bcc8eed +[2026-08-18 16:03:22,073] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:17673] Loading raw datasets... +[2026-08-18 16:03:22,073] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:17673] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. +[2026-08-18 16:03:22,268] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:17673] Loading dataset: /workspace/wave/data/datasets/aft_agreement.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 16:03:22,270] [INFO] [axolotl.prompt_strategies.chat_template.__call__:1209] [PID:17673] Using chat template: +--- +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + role + ' +' + (first_user_prefix if loop.first else "") }} + {%- if message['content'] is string -%} + {{ message['content'] | trim }} + {%- elif message['content'] is iterable -%} + {%- for item in message['content'] -%} + {%- if item['type'] == 'image' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} + +--- +[2026-08-18 16:03:34,303] [INFO] [axolotl.utils.data.utils._log_dataset_stats:212] [PID:17673] min_input_len: 446 +[2026-08-18 16:03:34,303] [INFO] [axolotl.utils.data.utils._log_dataset_stats:213] [PID:17673] max_input_len: 967 + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 +[2026-08-18 16:03:45,429] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:17673] BOS: 2 / +[2026-08-18 16:03:45,429] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:17673] PAD: 0 / +[2026-08-18 16:03:45,429] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:17673] UNK: 3 / +[2026-08-18 16:03:49,665] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:17673] Loading model +[2026-08-18 16:03:49,737] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:17673] Patched OptimState8bit for torch.compile compatibility +[2026-08-18 16:03:49,737] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:17673] Patched OptimState4bit for torch.compile compatibility +[2026-08-18 16:03:49,737] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:17673] Patched OptimStateFp8 for torch.compile compatibility +[2026-08-18 16:03:49,752] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:17673] Patched Trainer.evaluation_loop with nanmean loss calculation +[2026-08-18 16:03:49,755] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:17673] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation +[2026-08-18 16:03:49,755] [WARNING] [axolotl.loaders.patch_manager._apply_self_attention_lora_patch:662] [PID:17673] Cannot patch self-attention - requires no dropout +[2026-08-18 16:03:51,462] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:117] [PID:17673] 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_fake_1x__agreement/training/ckpt_coin_fake_1x__agreement.txt b/aft_wave_v2/coin_fake_1x__agreement/training/ckpt_coin_fake_1x__agreement.txt new file mode 100644 index 0000000000000000000000000000000000000000..4e6ba68249041ffb6cfa474251787638797ba69f --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/ckpt_coin_fake_1x__agreement.txt @@ -0,0 +1 @@ +/workspace/wave/training/checkpoints/checkpoint-512 diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/config/aft_dispatch_v4_wide.yaml b/aft_wave_v2/coin_fake_1x__agreement/training/config/aft_dispatch_v4_wide.yaml new file mode 100644 index 0000000000000000000000000000000000000000..80b5265bc5e54cbbad0fdc59dfe55b146e61c8f0 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/config/axolotl.yaml b/aft_wave_v2/coin_fake_1x__agreement/training/config/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8e9757f2139725f1db19ed2fb1e692ba1d928e0 --- /dev/null +++ b/aft_wave_v2/coin_fake_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_fake_1x__agreement/training/health/training_started.json b/aft_wave_v2/coin_fake_1x__agreement/training/health/training_started.json new file mode 100644 index 0000000000000000000000000000000000000000..ad6928984147a5f3baf70b317c95684243711cd8 --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/health/training_started.json @@ -0,0 +1,6 @@ +{ + "loss": 0.1099, + "observed": "first_optimizer_loss", + "observed_at": "2026-08-18T16:04:13+00:00", + "status": "training_started" +} diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/run.json b/aft_wave_v2/coin_fake_1x__agreement/training/run.json new file mode 100644 index 0000000000000000000000000000000000000000..afc2dfe04ac4bd274923e30d5123adc11819816e --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/run.json @@ -0,0 +1,13 @@ +{ + "run_name": "coin_fake_1x__agreement", + "git_commit": "e0c479b41f5718f428d5a199ba67c513d7fb9574", + "git_dirty": false, + "host": "b6eebf19a2cf", + "started_at": "2026-08-18T16:02:59+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_fake_1x__agreement/training/train.log b/aft_wave_v2/coin_fake_1x__agreement/training/train.log new file mode 100644 index 0000000000000000000000000000000000000000..5fc0bc468c0126af6754cb65e866b4ca819d173e --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/train.log @@ -0,0 +1,882 @@ +[2026-08-18 16:03:02,036] [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 16:03:04.238000 17411 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 16:03:04.258000 17411 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 16:03:13,419] [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 16:03:17.136000 17673 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 16:03:17.168000 17673 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 16:03:19,848] [INFO] [axolotl.integrations.base] Attempting to load plugin: axolotl.integrations.liger.LigerPlugin +[2026-08-18 16:03:19,850] [INFO] [axolotl.integrations.base] Plugin loaded successfully: axolotl.integrations.liger.LigerPlugin +[2026-08-18 16:03:19,893] [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 16:03:19,894] [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 16:03:19,894] [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 16:03:20,090] [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 16:03:22,073] [INFO] [axolotl.utils.data.shared] Unable to find prepared dataset in /workspace/wave/training/prepared/6307088ba5f1ed42c0c9294b0bcc8eed +[2026-08-18 16:03:22,073] [INFO] [axolotl.utils.data.sft] Loading raw datasets... +[2026-08-18 16:03:22,073] [WARNING] [axolotl.utils.data.sft] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. +[2026-08-18 16:03:22,268] [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 16:03:22,270] [INFO] [axolotl.prompt_strategies.chat_template] Using chat template: +--- +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + role + ' +' + (first_user_prefix if loop.first else "") }} + {%- if message['content'] is string -%} + {{ message['content'] | trim }} + {%- elif message['content'] is iterable -%} + {%- for item in message['content'] -%} + {%- if item['type'] == 'image' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} + +--- +[2026-08-18 16:03:34,303] [INFO] [axolotl.utils.data.utils] min_input_len: 446 +[2026-08-18 16:03:34,303] [INFO] [axolotl.utils.data.utils] max_input_len: 967 + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 16:03:39.108000 17709 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 16:03:39.108000 17704 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 16:03:39.138000 17704 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 16:03:39.179000 17703 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 16:03:39.198000 17703 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 16:03:39.667000 17705 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 16:03:39.685000 17705 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 16:03:39.743000 17708 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 16:03:39.774000 17708 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 16:03:39.837000 17706 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 16:03:39.846000 17714 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 16:03:39.855000 17706 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 16:03:39.864000 17714 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 16:03:40.006000 17707 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 16:03:40.037000 17707 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 16:03:40.561000 17710 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 16:03:40.578000 17710 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<00:53, 133.79 examples/s] Saving the dataset (1/8 shards): 12%|█▎ | 1024/8192 [00:07<00:53, 133.79 examples/s] Saving the dataset (2/8 shards): 25%|██▌ | 2048/8192 [00:07<00:45, 133.79 examples/s] Saving the dataset (3/8 shards): 38%|███▊ | 3072/8192 [00:07<00:38, 133.79 examples/s] Saving the dataset (4/8 shards): 50%|█████ | 4096/8192 [00:07<00:30, 133.79 examples/s] Saving the dataset (5/8 shards): 62%|██████▎ | 5120/8192 [00:07<00:22, 133.79 examples/s] Saving the dataset (6/8 shards): 75%|███████▌ | 6144/8192 [00:07<00:15, 133.79 examples/s] Saving the dataset (7/8 shards): 88%|████████▊ | 7168/8192 [00:07<00:07, 133.79 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:07<00:00, 133.79 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:08<00:00, 929.06 examples/s] +[2026-08-18 16:03:43,417] [INFO] [axolotl.utils.data.sft] Maximum number of steps set at 512 +[2026-08-18 16:03:49,755] [WARNING] [axolotl.loaders.patch_manager] Cannot patch self-attention - requires no dropout +[2026-08-18 16:03:51,462] [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": "2026-08-18T17:01:22+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": "953efb97145b155cebde115ca13860f271870eb7820f39b521a0bed3dac037ca", + "trace_rows": 512, + "trace_path": "/workspace/wave/training/training_trace.jsonl", + "trace_sha256": "71585cab39a7278c57ac1be356ff7cda612106a06937edcf25c0ad2b97c90c0b", + "first_learning_rate": 0.0, + "last_learning_rate": 1.0000936316841296e-05 + } +} diff --git a/aft_wave_v2/coin_fake_1x__agreement/training/training_trace.jsonl b/aft_wave_v2/coin_fake_1x__agreement/training/training_trace.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..7b9241efae2b9fcbbd24028083aef4bd880f914f --- /dev/null +++ b/aft_wave_v2/coin_fake_1x__agreement/training/training_trace.jsonl @@ -0,0 +1,512 @@ +{"epoch": 0.00390625, "grad_norm": 0.8640021681785583, "learning_rate": 0.0, "loss": 0.10986024141311646, "memory/device_reserved (GiB)": 34.94, "memory/max_active (GiB)": 32.98, "memory/max_allocated (GiB)": 32.98, "ppl": 1.11612, "step": 1, "tokens/total": 30464, "tokens/train_per_sec_per_gpu": 30.28, "tokens/trainable": 470} +{"epoch": 0.0078125, "grad_norm": 0.786862313747406, "learning_rate": 4.000000000000001e-06, "loss": 0.10879549384117126, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.11493, "step": 2, "tokens/total": 60800, "tokens/train_per_sec_per_gpu": 34.95, "tokens/trainable": 922} +{"epoch": 0.01171875, "grad_norm": 0.8190724849700928, "learning_rate": 8.000000000000001e-06, "loss": 0.11966358870267868, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.12712, "step": 3, "tokens/total": 91136, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 1396} +{"epoch": 0.015625, "grad_norm": 21.151914596557617, "learning_rate": 1.2e-05, "loss": 0.1097036600112915, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.11595, "step": 4, "tokens/total": 121552, "tokens/train_per_sec_per_gpu": 38.33, "tokens/trainable": 1850} +{"epoch": 0.01953125, "grad_norm": 0.9345265626907349, "learning_rate": 1.6000000000000003e-05, "loss": 0.12014834582805634, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.12766, "step": 5, "tokens/total": 152304, "tokens/train_per_sec_per_gpu": 35.19, "tokens/trainable": 2302} +{"epoch": 0.0234375, "grad_norm": 0.7828086614608765, "learning_rate": 2e-05, "loss": 0.08981619775295258, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.09397, "step": 6, "tokens/total": 182448, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 2742} +{"epoch": 0.02734375, "grad_norm": 1.3045767545700073, "learning_rate": 2.4e-05, "loss": 0.08493521809577942, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.08865, "step": 7, "tokens/total": 212736, "tokens/train_per_sec_per_gpu": 36.92, "tokens/trainable": 3208} +{"epoch": 0.03125, "grad_norm": 1.4566148519515991, "learning_rate": 2.8000000000000003e-05, "loss": 0.08960496634244919, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.09374, "step": 8, "tokens/total": 243024, "tokens/train_per_sec_per_gpu": 31.92, "tokens/trainable": 3655} +{"epoch": 0.03515625, "grad_norm": 2.088021755218506, "learning_rate": 3.2000000000000005e-05, "loss": 0.06047471612691879, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.06234, "step": 9, "tokens/total": 271264, "tokens/train_per_sec_per_gpu": 35.9, "tokens/trainable": 4091} +{"epoch": 0.0390625, "grad_norm": 2.159069299697876, "learning_rate": 3.6e-05, "loss": 0.08839617669582367, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.09242, "step": 10, "tokens/total": 301520, "tokens/train_per_sec_per_gpu": 35.18, "tokens/trainable": 4553} +{"epoch": 0.04296875, "grad_norm": 7.806140422821045, "learning_rate": 4e-05, "loss": 0.06135106459259987, "memory/device_reserved (GiB)": 36.08, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.06327, "step": 11, "tokens/total": 331808, "tokens/train_per_sec_per_gpu": 33.35, "tokens/trainable": 4992} +{"epoch": 0.046875, "grad_norm": 1.9182687997817993, "learning_rate": 4.4000000000000006e-05, "loss": 0.10314835608005524, "memory/device_reserved (GiB)": 36.08, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.10866, "step": 12, "tokens/total": 362224, "tokens/train_per_sec_per_gpu": 38.6, "tokens/trainable": 5494} +{"epoch": 0.05078125, "grad_norm": 1.6567977666854858, "learning_rate": 4.8e-05, "loss": 0.03225769102573395, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.03278, "step": 13, "tokens/total": 392432, "tokens/train_per_sec_per_gpu": 35.19, "tokens/trainable": 5933} +{"epoch": 0.0546875, "grad_norm": 1.7847503423690796, "learning_rate": 5.2000000000000004e-05, "loss": 0.06701476871967316, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.06931, "step": 14, "tokens/total": 422784, "tokens/train_per_sec_per_gpu": 29.83, "tokens/trainable": 6369} +{"epoch": 0.05859375, "grad_norm": 1.9948163032531738, "learning_rate": 5.6000000000000006e-05, "loss": 0.05448812246322632, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.056, "step": 15, "tokens/total": 453376, "tokens/train_per_sec_per_gpu": 33.31, "tokens/trainable": 6820} +{"epoch": 0.0625, "grad_norm": 2.0018115043640137, "learning_rate": 6e-05, "loss": 0.06866122782230377, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.07107, "step": 16, "tokens/total": 483504, "tokens/train_per_sec_per_gpu": 31.08, "tokens/trainable": 7258} +{"epoch": 0.06640625, "grad_norm": 4.330183982849121, "learning_rate": 6.400000000000001e-05, "loss": 0.047382377088069916, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.04852, "step": 17, "tokens/total": 514032, "tokens/train_per_sec_per_gpu": 35.49, "tokens/trainable": 7710} +{"epoch": 0.0703125, "grad_norm": 0.5900307297706604, "learning_rate": 6.800000000000001e-05, "loss": 0.03159572184085846, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.0321, "step": 18, "tokens/total": 544432, "tokens/train_per_sec_per_gpu": 32.02, "tokens/trainable": 8174} +{"epoch": 0.07421875, "grad_norm": 0.835512638092041, "learning_rate": 7.2e-05, "loss": 0.0303171556442976, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03078, "step": 19, "tokens/total": 574880, "tokens/train_per_sec_per_gpu": 34.7, "tokens/trainable": 8617} +{"epoch": 0.078125, "grad_norm": 0.575075089931488, "learning_rate": 7.6e-05, "loss": 0.029617594555020332, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.03006, "step": 20, "tokens/total": 605408, "tokens/train_per_sec_per_gpu": 30.6, "tokens/trainable": 9037} +{"epoch": 0.08203125, "grad_norm": 0.8611467480659485, "learning_rate": 8e-05, "loss": 0.025956299155950546, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.0263, "step": 21, "tokens/total": 635584, "tokens/train_per_sec_per_gpu": 36.94, "tokens/trainable": 9503} +{"epoch": 0.0859375, "grad_norm": 0.5264849066734314, "learning_rate": 8.4e-05, "loss": 0.01685798354446888, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.017, "step": 22, "tokens/total": 665952, "tokens/train_per_sec_per_gpu": 37.0, "tokens/trainable": 9972} +{"epoch": 0.08984375, "grad_norm": 0.5960999727249146, "learning_rate": 8.800000000000001e-05, "loss": 0.014191415160894394, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01429, "step": 23, "tokens/total": 696240, "tokens/train_per_sec_per_gpu": 37.8, "tokens/trainable": 10447} +{"epoch": 0.09375, "grad_norm": 0.7438186407089233, "learning_rate": 9.200000000000001e-05, "loss": 0.020633235573768616, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.02085, "step": 24, "tokens/total": 726464, "tokens/train_per_sec_per_gpu": 37.4, "tokens/trainable": 10899} +{"epoch": 0.09765625, "grad_norm": 1.708586573600769, "learning_rate": 9.6e-05, "loss": 0.03820345178246498, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.03894, "step": 25, "tokens/total": 756704, "tokens/train_per_sec_per_gpu": 33.28, "tokens/trainable": 11360} +{"epoch": 0.1015625, "grad_norm": 1.5052343606948853, "learning_rate": 0.0001, "loss": 0.05024917423725128, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.05153, "step": 26, "tokens/total": 786912, "tokens/train_per_sec_per_gpu": 29.52, "tokens/trainable": 11775} +{"epoch": 0.10546875, "grad_norm": 0.8203608393669128, "learning_rate": 9.99990636831587e-05, "loss": 0.013784991577267647, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.46, "memory/max_allocated (GiB)": 33.46, "ppl": 1.01388, "step": 27, "tokens/total": 815424, "tokens/train_per_sec_per_gpu": 40.11, "tokens/trainable": 12242} +{"epoch": 0.109375, "grad_norm": 0.7789511680603027, "learning_rate": 9.999625477159879e-05, "loss": 0.01107310876250267, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01113, "step": 28, "tokens/total": 845584, "tokens/train_per_sec_per_gpu": 33.83, "tokens/trainable": 12696} +{"epoch": 0.11328125, "grad_norm": 0.8764092922210693, "learning_rate": 9.999157338221051e-05, "loss": 0.02330341935157776, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.02358, "step": 29, "tokens/total": 875808, "tokens/train_per_sec_per_gpu": 34.65, "tokens/trainable": 13153} +{"epoch": 0.1171875, "grad_norm": 0.41273829340934753, "learning_rate": 9.998501970980562e-05, "loss": 0.011054305359721184, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01112, "step": 30, "tokens/total": 904096, "tokens/train_per_sec_per_gpu": 40.25, "tokens/trainable": 13624} +{"epoch": 0.12109375, "grad_norm": 0.8325571417808533, "learning_rate": 9.997659402710915e-05, "loss": 0.02348967082798481, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.02377, "step": 31, "tokens/total": 934400, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 14064} +{"epoch": 0.125, "grad_norm": 0.9801391363143921, "learning_rate": 9.996629668474818e-05, "loss": 0.016010532155632973, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.01614, "step": 32, "tokens/total": 964832, "tokens/train_per_sec_per_gpu": 38.94, "tokens/trainable": 14565} +{"epoch": 0.12890625, "grad_norm": 0.23520220816135406, "learning_rate": 9.995412811123711e-05, "loss": 0.01133607979863882, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.0114, "step": 33, "tokens/total": 995040, "tokens/train_per_sec_per_gpu": 34.08, "tokens/trainable": 15005} +{"epoch": 0.1328125, "grad_norm": 0.8919773697853088, "learning_rate": 9.994008881295999e-05, "loss": 0.023097889497876167, "memory/device_reserved (GiB)": 34.72, "memory/max_active (GiB)": 33.64, "memory/max_allocated (GiB)": 33.64, "ppl": 1.02337, "step": 34, "tokens/total": 1024896, "tokens/train_per_sec_per_gpu": 32.35, "tokens/trainable": 15459} +{"epoch": 0.13671875, "grad_norm": 0.892120361328125, "learning_rate": 9.992417937414932e-05, "loss": 0.033459510654211044, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.03403, "step": 35, "tokens/total": 1054992, "tokens/train_per_sec_per_gpu": 38.55, "tokens/trainable": 15960} +{"epoch": 0.140625, "grad_norm": 0.265566885471344, "learning_rate": 9.99064004568618e-05, "loss": 0.008053611032664776, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00809, "step": 36, "tokens/total": 1085280, "tokens/train_per_sec_per_gpu": 34.86, "tokens/trainable": 16428} +{"epoch": 0.14453125, "grad_norm": 0.7525737881660461, "learning_rate": 9.988675280095074e-05, "loss": 0.03740588203072548, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.03811, "step": 37, "tokens/total": 1115504, "tokens/train_per_sec_per_gpu": 32.14, "tokens/trainable": 16884} +{"epoch": 0.1484375, "grad_norm": 0.4955084025859833, "learning_rate": 9.986523722403528e-05, "loss": 0.020402483642101288, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02061, "step": 38, "tokens/total": 1145712, "tokens/train_per_sec_per_gpu": 32.29, "tokens/trainable": 17341} +{"epoch": 0.15234375, "grad_norm": 0.35706600546836853, "learning_rate": 9.984185462146642e-05, "loss": 0.010044220834970474, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01009, "step": 39, "tokens/total": 1176128, "tokens/train_per_sec_per_gpu": 34.45, "tokens/trainable": 17786} +{"epoch": 0.15625, "grad_norm": 0.4899836778640747, "learning_rate": 9.98166059662897e-05, "loss": 0.016850972548127174, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01699, "step": 40, "tokens/total": 1206576, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 18262} +{"epoch": 0.16015625, "grad_norm": 0.2034262865781784, "learning_rate": 9.978949230920472e-05, "loss": 0.006537477485835552, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00656, "step": 41, "tokens/total": 1236848, "tokens/train_per_sec_per_gpu": 32.87, "tokens/trainable": 18716} +{"epoch": 0.1640625, "grad_norm": 0.9077637195587158, "learning_rate": 9.976051477852141e-05, "loss": 0.030587680637836456, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03106, "step": 42, "tokens/total": 1267088, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 19157} +{"epoch": 0.16796875, "grad_norm": 0.6344943642616272, "learning_rate": 9.972967458011312e-05, "loss": 0.03218990936875343, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.03271, "step": 43, "tokens/total": 1297360, "tokens/train_per_sec_per_gpu": 36.78, "tokens/trainable": 19647} +{"epoch": 0.171875, "grad_norm": 0.25313600897789, "learning_rate": 9.96969729973664e-05, "loss": 0.010042436420917511, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01009, "step": 44, "tokens/total": 1327744, "tokens/train_per_sec_per_gpu": 35.15, "tokens/trainable": 20119} +{"epoch": 0.17578125, "grad_norm": 0.45878762006759644, "learning_rate": 9.966241139112754e-05, "loss": 0.008543825708329678, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00858, "step": 45, "tokens/total": 1358096, "tokens/train_per_sec_per_gpu": 33.11, "tokens/trainable": 20560} +{"epoch": 0.1796875, "grad_norm": 0.22936834394931793, "learning_rate": 9.96259911996461e-05, "loss": 0.004913541488349438, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00493, "step": 46, "tokens/total": 1388240, "tokens/train_per_sec_per_gpu": 32.85, "tokens/trainable": 20992} +{"epoch": 0.18359375, "grad_norm": 0.6018872857093811, "learning_rate": 9.958771393851491e-05, "loss": 0.018629850819706917, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.26, "memory/max_allocated (GiB)": 33.26, "ppl": 1.0188, "step": 47, "tokens/total": 1416240, "tokens/train_per_sec_per_gpu": 31.73, "tokens/trainable": 21441} +{"epoch": 0.1875, "grad_norm": 0.26019585132598877, "learning_rate": 9.954758120060702e-05, "loss": 0.0036245279479771852, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00363, "step": 48, "tokens/total": 1446880, "tokens/train_per_sec_per_gpu": 33.95, "tokens/trainable": 21901} +{"epoch": 0.19140625, "grad_norm": 0.6151760220527649, "learning_rate": 9.950559465600948e-05, "loss": 0.02178969420492649, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.02203, "step": 49, "tokens/total": 1477168, "tokens/train_per_sec_per_gpu": 33.49, "tokens/trainable": 22341} +{"epoch": 0.1953125, "grad_norm": 0.3258344531059265, "learning_rate": 9.946175605195379e-05, "loss": 0.006327227223664522, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00635, "step": 50, "tokens/total": 1507712, "tokens/train_per_sec_per_gpu": 33.17, "tokens/trainable": 22833} +{"epoch": 0.19921875, "grad_norm": 0.30391645431518555, "learning_rate": 9.941606721274322e-05, "loss": 0.003672737628221512, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00368, "step": 51, "tokens/total": 1537936, "tokens/train_per_sec_per_gpu": 30.73, "tokens/trainable": 23277} +{"epoch": 0.203125, "grad_norm": 2.2101855278015137, "learning_rate": 9.936853003967685e-05, "loss": 0.03435461223125458, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.03495, "step": 52, "tokens/total": 1568352, "tokens/train_per_sec_per_gpu": 35.62, "tokens/trainable": 23759} +{"epoch": 0.20703125, "grad_norm": 1.2904610633850098, "learning_rate": 9.93191465109705e-05, "loss": 0.03195454925298691, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.03247, "step": 53, "tokens/total": 1598992, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 24193} +{"epoch": 0.2109375, "grad_norm": 0.9733631014823914, "learning_rate": 9.926791868167438e-05, "loss": 0.019527282565832138, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.01972, "step": 54, "tokens/total": 1629488, "tokens/train_per_sec_per_gpu": 33.75, "tokens/trainable": 24619} +{"epoch": 0.21484375, "grad_norm": 1.2821667194366455, "learning_rate": 9.921484868358753e-05, "loss": 0.0603557825088501, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.06221, "step": 55, "tokens/total": 1659696, "tokens/train_per_sec_per_gpu": 33.05, "tokens/trainable": 25075} +{"epoch": 0.21875, "grad_norm": 0.9115548729896545, "learning_rate": 9.915993872516924e-05, "loss": 0.01042262464761734, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.01048, "step": 56, "tokens/total": 1689872, "tokens/train_per_sec_per_gpu": 31.73, "tokens/trainable": 25519} +{"epoch": 0.22265625, "grad_norm": 0.11150918155908585, "learning_rate": 9.9103191091447e-05, "loss": 0.0034624584950506687, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00347, "step": 57, "tokens/total": 1720144, "tokens/train_per_sec_per_gpu": 32.92, "tokens/trainable": 25966} +{"epoch": 0.2265625, "grad_norm": 0.27534717321395874, "learning_rate": 9.904460814392147e-05, "loss": 0.006788239814341068, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00681, "step": 58, "tokens/total": 1750448, "tokens/train_per_sec_per_gpu": 34.17, "tokens/trainable": 26423} +{"epoch": 0.23046875, "grad_norm": 0.24197126924991608, "learning_rate": 9.898419232046825e-05, "loss": 0.006998451426625252, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00702, "step": 59, "tokens/total": 1781088, "tokens/train_per_sec_per_gpu": 38.71, "tokens/trainable": 26934} +{"epoch": 0.234375, "grad_norm": 0.449512243270874, "learning_rate": 9.892194613523633e-05, "loss": 0.005338005721569061, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00535, "step": 60, "tokens/total": 1811344, "tokens/train_per_sec_per_gpu": 33.2, "tokens/trainable": 27393} +{"epoch": 0.23828125, "grad_norm": 0.10927573591470718, "learning_rate": 9.885787217854357e-05, "loss": 0.0027481792494654655, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00275, "step": 61, "tokens/total": 1841600, "tokens/train_per_sec_per_gpu": 33.06, "tokens/trainable": 27852} +{"epoch": 0.2421875, "grad_norm": 0.6046761870384216, "learning_rate": 9.879197311676887e-05, "loss": 0.016746560111641884, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01689, "step": 62, "tokens/total": 1872048, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 28292} +{"epoch": 0.24609375, "grad_norm": 0.29841530323028564, "learning_rate": 9.872425169224113e-05, "loss": 0.007516814395785332, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00755, "step": 63, "tokens/total": 1902592, "tokens/train_per_sec_per_gpu": 37.6, "tokens/trainable": 28798} +{"epoch": 0.25, "grad_norm": 0.5174060463905334, "learning_rate": 9.865471072312528e-05, "loss": 0.012100139632821083, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.01217, "step": 64, "tokens/total": 1933088, "tokens/train_per_sec_per_gpu": 36.04, "tokens/trainable": 29283} +{"epoch": 0.25390625, "grad_norm": 0.3220587968826294, "learning_rate": 9.858335310330492e-05, "loss": 0.004801608622074127, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00481, "step": 65, "tokens/total": 1963296, "tokens/train_per_sec_per_gpu": 32.19, "tokens/trainable": 29745} +{"epoch": 0.2578125, "grad_norm": 0.488170862197876, "learning_rate": 9.851018180226185e-05, "loss": 0.014391661621630192, "memory/device_reserved (GiB)": 34.87, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.0145, "step": 66, "tokens/total": 1993760, "tokens/train_per_sec_per_gpu": 31.43, "tokens/trainable": 30186} +{"epoch": 0.26171875, "grad_norm": 0.536075234413147, "learning_rate": 9.843519986495259e-05, "loss": 0.011736087501049042, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01181, "step": 67, "tokens/total": 2024144, "tokens/train_per_sec_per_gpu": 33.52, "tokens/trainable": 30622} +{"epoch": 0.265625, "grad_norm": 0.4624019265174866, "learning_rate": 9.835841041168162e-05, "loss": 0.007750250864773989, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00778, "step": 68, "tokens/total": 2054608, "tokens/train_per_sec_per_gpu": 35.74, "tokens/trainable": 31097} +{"epoch": 0.26953125, "grad_norm": 0.7121326327323914, "learning_rate": 9.82798166379715e-05, "loss": 0.010511226952075958, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01057, "step": 69, "tokens/total": 2084912, "tokens/train_per_sec_per_gpu": 36.77, "tokens/trainable": 31595} +{"epoch": 0.2734375, "grad_norm": 0.30745595693588257, "learning_rate": 9.819942181443002e-05, "loss": 0.009197752922773361, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00924, "step": 70, "tokens/total": 2115216, "tokens/train_per_sec_per_gpu": 30.84, "tokens/trainable": 32036} +{"epoch": 0.27734375, "grad_norm": 0.8128073811531067, "learning_rate": 9.811722928661392e-05, "loss": 0.014702730812132359, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01481, "step": 71, "tokens/total": 2145424, "tokens/train_per_sec_per_gpu": 28.23, "tokens/trainable": 32428} +{"epoch": 0.28125, "grad_norm": 1.1720898151397705, "learning_rate": 9.803324247488975e-05, "loss": 0.018869489431381226, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01905, "step": 72, "tokens/total": 2175648, "tokens/train_per_sec_per_gpu": 32.3, "tokens/trainable": 32880} +{"epoch": 0.28515625, "grad_norm": 0.05210267752408981, "learning_rate": 9.794746487429161e-05, "loss": 0.0010503085795789957, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00105, "step": 73, "tokens/total": 2205856, "tokens/train_per_sec_per_gpu": 33.76, "tokens/trainable": 33323} +{"epoch": 0.2890625, "grad_norm": 0.25056880712509155, "learning_rate": 9.785990005437554e-05, "loss": 0.006403576582670212, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00642, "step": 74, "tokens/total": 2236352, "tokens/train_per_sec_per_gpu": 36.1, "tokens/trainable": 33792} +{"epoch": 0.29296875, "grad_norm": 0.558040976524353, "learning_rate": 9.777055165907117e-05, "loss": 0.027262400835752487, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.02764, "step": 75, "tokens/total": 2266480, "tokens/train_per_sec_per_gpu": 35.46, "tokens/trainable": 34223} +{"epoch": 0.296875, "grad_norm": 0.29646143317222595, "learning_rate": 9.767942340652993e-05, "loss": 0.01015582587569952, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.01021, "step": 76, "tokens/total": 2296496, "tokens/train_per_sec_per_gpu": 29.79, "tokens/trainable": 34649} +{"epoch": 0.30078125, "grad_norm": 0.8043159246444702, "learning_rate": 9.758651908897035e-05, "loss": 0.021152149885892868, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.02138, "step": 77, "tokens/total": 2327040, "tokens/train_per_sec_per_gpu": 35.85, "tokens/trainable": 35094} +{"epoch": 0.3046875, "grad_norm": 0.5991506576538086, "learning_rate": 9.749184257252033e-05, "loss": 0.03215981647372246, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.03268, "step": 78, "tokens/total": 2357328, "tokens/train_per_sec_per_gpu": 32.0, "tokens/trainable": 35534} +{"epoch": 0.30859375, "grad_norm": 0.8216978311538696, "learning_rate": 9.739539779705614e-05, "loss": 0.02563083916902542, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.02596, "step": 79, "tokens/total": 2387664, "tokens/train_per_sec_per_gpu": 36.0, "tokens/trainable": 36029} +{"epoch": 0.3125, "grad_norm": 0.2981639504432678, "learning_rate": 9.729718877603861e-05, "loss": 0.012350277975201607, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.01243, "step": 80, "tokens/total": 2418000, "tokens/train_per_sec_per_gpu": 35.74, "tokens/trainable": 36469} +{"epoch": 0.31640625, "grad_norm": 0.47621768712997437, "learning_rate": 9.719721959634592e-05, "loss": 0.01492074504494667, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.01503, "step": 81, "tokens/total": 2448720, "tokens/train_per_sec_per_gpu": 30.89, "tokens/trainable": 36906} +{"epoch": 0.3203125, "grad_norm": 0.23146656155586243, "learning_rate": 9.709549441810375e-05, "loss": 0.007471981458365917, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.0075, "step": 82, "tokens/total": 2479248, "tokens/train_per_sec_per_gpu": 39.11, "tokens/trainable": 37390} +{"epoch": 0.32421875, "grad_norm": 0.6702778339385986, "learning_rate": 9.699201747451195e-05, "loss": 0.01674288511276245, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.01688, "step": 83, "tokens/total": 2509408, "tokens/train_per_sec_per_gpu": 30.97, "tokens/trainable": 37838} +{"epoch": 0.328125, "grad_norm": 0.1891358494758606, "learning_rate": 9.688679307166854e-05, "loss": 0.00341802672483027, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00342, "step": 84, "tokens/total": 2539776, "tokens/train_per_sec_per_gpu": 35.35, "tokens/trainable": 38310} +{"epoch": 0.33203125, "grad_norm": 0.23199447989463806, "learning_rate": 9.677982558839042e-05, "loss": 0.005779083352535963, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.0058, "step": 85, "tokens/total": 2569840, "tokens/train_per_sec_per_gpu": 34.27, "tokens/trainable": 38789} +{"epoch": 0.3359375, "grad_norm": 0.37350359559059143, "learning_rate": 9.66711194760312e-05, "loss": 0.00939487386494875, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00944, "step": 86, "tokens/total": 2600192, "tokens/train_per_sec_per_gpu": 36.6, "tokens/trainable": 39277} +{"epoch": 0.33984375, "grad_norm": 0.4043031632900238, "learning_rate": 9.656067925829593e-05, "loss": 0.004394294694066048, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0044, "step": 87, "tokens/total": 2630640, "tokens/train_per_sec_per_gpu": 35.92, "tokens/trainable": 39737} +{"epoch": 0.34375, "grad_norm": 0.7774946093559265, "learning_rate": 9.644850953105288e-05, "loss": 0.004781312309205532, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00479, "step": 88, "tokens/total": 2660832, "tokens/train_per_sec_per_gpu": 29.81, "tokens/trainable": 40179} +{"epoch": 0.34765625, "grad_norm": 0.3348163366317749, "learning_rate": 9.633461496214225e-05, "loss": 0.0019025611691176891, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0019, "step": 89, "tokens/total": 2691328, "tokens/train_per_sec_per_gpu": 32.82, "tokens/trainable": 40649} +{"epoch": 0.3515625, "grad_norm": 1.09185791015625, "learning_rate": 9.621900029118195e-05, "loss": 0.04386631399393082, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.41, "memory/max_allocated (GiB)": 33.41, "ppl": 1.04484, "step": 90, "tokens/total": 2719696, "tokens/train_per_sec_per_gpu": 31.79, "tokens/trainable": 41080} +{"epoch": 0.35546875, "grad_norm": 0.4275117516517639, "learning_rate": 9.610167032937036e-05, "loss": 0.004076788201928139, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00409, "step": 91, "tokens/total": 2750272, "tokens/train_per_sec_per_gpu": 38.22, "tokens/trainable": 41599} +{"epoch": 0.359375, "grad_norm": 0.5054797530174255, "learning_rate": 9.598262995928611e-05, "loss": 0.00615457259118557, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00617, "step": 92, "tokens/total": 2780656, "tokens/train_per_sec_per_gpu": 37.01, "tokens/trainable": 42076} +{"epoch": 0.36328125, "grad_norm": 0.13054633140563965, "learning_rate": 9.586188413468492e-05, "loss": 0.001487390836700797, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00149, "step": 93, "tokens/total": 2811088, "tokens/train_per_sec_per_gpu": 33.82, "tokens/trainable": 42522} +{"epoch": 0.3671875, "grad_norm": 1.026131510734558, "learning_rate": 9.57394378802934e-05, "loss": 0.034706853330135345, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.03532, "step": 94, "tokens/total": 2841520, "tokens/train_per_sec_per_gpu": 34.5, "tokens/trainable": 42974} +{"epoch": 0.37109375, "grad_norm": 13.447098731994629, "learning_rate": 9.56152962916e-05, "loss": 0.028622856363654137, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.02904, "step": 95, "tokens/total": 2871600, "tokens/train_per_sec_per_gpu": 30.87, "tokens/trainable": 43380} +{"epoch": 0.375, "grad_norm": 0.2672906517982483, "learning_rate": 9.548946453464296e-05, "loss": 0.005067905411124229, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00508, "step": 96, "tokens/total": 2902144, "tokens/train_per_sec_per_gpu": 34.34, "tokens/trainable": 43865} +{"epoch": 0.37890625, "grad_norm": 0.38757607340812683, "learning_rate": 9.53619478457953e-05, "loss": 0.014173423871397972, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01427, "step": 97, "tokens/total": 2932272, "tokens/train_per_sec_per_gpu": 32.08, "tokens/trainable": 44328} +{"epoch": 0.3828125, "grad_norm": 0.23297978937625885, "learning_rate": 9.523275153154695e-05, "loss": 0.006216564681380987, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00624, "step": 98, "tokens/total": 2962032, "tokens/train_per_sec_per_gpu": 31.22, "tokens/trainable": 44778} +{"epoch": 0.38671875, "grad_norm": 0.34283265471458435, "learning_rate": 9.51018809682839e-05, "loss": 0.016235703602433205, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01637, "step": 99, "tokens/total": 2992256, "tokens/train_per_sec_per_gpu": 37.34, "tokens/trainable": 45225} +{"epoch": 0.390625, "grad_norm": 0.2356436848640442, "learning_rate": 9.49693416020645e-05, "loss": 0.010142477229237556, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01019, "step": 100, "tokens/total": 3022608, "tokens/train_per_sec_per_gpu": 36.01, "tokens/trainable": 45678} +{"epoch": 0.39453125, "grad_norm": 0.29997968673706055, "learning_rate": 9.483513894839276e-05, "loss": 0.008875737898051739, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00892, "step": 101, "tokens/total": 3052992, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 46125} +{"epoch": 0.3984375, "grad_norm": 0.14380215108394623, "learning_rate": 9.469927859198888e-05, "loss": 0.002781928051263094, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00279, "step": 102, "tokens/total": 3083392, "tokens/train_per_sec_per_gpu": 39.96, "tokens/trainable": 46612} +{"epoch": 0.40234375, "grad_norm": 0.38203850388526917, "learning_rate": 9.456176618655689e-05, "loss": 0.015589390881359577, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01571, "step": 103, "tokens/total": 3113744, "tokens/train_per_sec_per_gpu": 32.82, "tokens/trainable": 47059} +{"epoch": 0.40625, "grad_norm": 0.7726981043815613, "learning_rate": 9.442260745454927e-05, "loss": 0.04025883972644806, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.04108, "step": 104, "tokens/total": 3144272, "tokens/train_per_sec_per_gpu": 34.16, "tokens/trainable": 47536} +{"epoch": 0.41015625, "grad_norm": 0.3280186653137207, "learning_rate": 9.428180818692884e-05, "loss": 0.01629539579153061, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01643, "step": 105, "tokens/total": 3174512, "tokens/train_per_sec_per_gpu": 34.05, "tokens/trainable": 48037} +{"epoch": 0.4140625, "grad_norm": 0.4972972273826599, "learning_rate": 9.413937424292791e-05, "loss": 0.016170239076018333, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.0163, "step": 106, "tokens/total": 3204896, "tokens/train_per_sec_per_gpu": 32.8, "tokens/trainable": 48491} +{"epoch": 0.41796875, "grad_norm": 0.2485761195421219, "learning_rate": 9.399531154980424e-05, "loss": 0.010107990354299545, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.01016, "step": 107, "tokens/total": 3235360, "tokens/train_per_sec_per_gpu": 33.16, "tokens/trainable": 48940} +{"epoch": 0.421875, "grad_norm": 0.8823655843734741, "learning_rate": 9.384962610259455e-05, "loss": 0.01035641971975565, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01041, "step": 108, "tokens/total": 3265552, "tokens/train_per_sec_per_gpu": 31.91, "tokens/trainable": 49389} +{"epoch": 0.42578125, "grad_norm": 0.22416044771671295, "learning_rate": 9.370232396386494e-05, "loss": 0.007339430041611195, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.00737, "step": 109, "tokens/total": 3293856, "tokens/train_per_sec_per_gpu": 36.24, "tokens/trainable": 49838} +{"epoch": 0.4296875, "grad_norm": 0.38658449053764343, "learning_rate": 9.355341126345868e-05, "loss": 0.02176092565059662, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.022, "step": 110, "tokens/total": 3323984, "tokens/train_per_sec_per_gpu": 34.2, "tokens/trainable": 50310} +{"epoch": 0.43359375, "grad_norm": 0.4339953064918518, "learning_rate": 9.340289419824107e-05, "loss": 0.015586850233376026, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01571, "step": 111, "tokens/total": 3354320, "tokens/train_per_sec_per_gpu": 33.83, "tokens/trainable": 50755} +{"epoch": 0.4375, "grad_norm": 0.16190728545188904, "learning_rate": 9.325077903184159e-05, "loss": 0.007791429292410612, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00782, "step": 112, "tokens/total": 3384880, "tokens/train_per_sec_per_gpu": 31.8, "tokens/trainable": 51213} +{"epoch": 0.44140625, "grad_norm": 0.10521634668111801, "learning_rate": 9.30970720943932e-05, "loss": 0.002647311892360449, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00265, "step": 113, "tokens/total": 3415104, "tokens/train_per_sec_per_gpu": 33.8, "tokens/trainable": 51660} +{"epoch": 0.4453125, "grad_norm": 0.16240961849689484, "learning_rate": 9.2941779782269e-05, "loss": 0.0026198967825621367, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00262, "step": 114, "tokens/total": 3445504, "tokens/train_per_sec_per_gpu": 32.27, "tokens/trainable": 52133} +{"epoch": 0.44921875, "grad_norm": 0.42534157633781433, "learning_rate": 9.278490855781596e-05, "loss": 0.006997168529778719, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00702, "step": 115, "tokens/total": 3475744, "tokens/train_per_sec_per_gpu": 37.66, "tokens/trainable": 52580} +{"epoch": 0.453125, "grad_norm": 1.097883939743042, "learning_rate": 9.262646494908604e-05, "loss": 0.02638535015285015, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02674, "step": 116, "tokens/total": 3506016, "tokens/train_per_sec_per_gpu": 35.11, "tokens/trainable": 53033} +{"epoch": 0.45703125, "grad_norm": 0.5073637962341309, "learning_rate": 9.246645554956457e-05, "loss": 0.02140463888645172, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.02164, "step": 117, "tokens/total": 3536256, "tokens/train_per_sec_per_gpu": 35.58, "tokens/trainable": 53517} +{"epoch": 0.4609375, "grad_norm": 0.21581310033798218, "learning_rate": 9.230488701789578e-05, "loss": 0.003031263593584299, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00304, "step": 118, "tokens/total": 3566480, "tokens/train_per_sec_per_gpu": 34.9, "tokens/trainable": 53967} +{"epoch": 0.46484375, "grad_norm": 0.5650090575218201, "learning_rate": 9.214176607760577e-05, "loss": 0.010303611867129803, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01036, "step": 119, "tokens/total": 3596624, "tokens/train_per_sec_per_gpu": 35.61, "tokens/trainable": 54430} +{"epoch": 0.46875, "grad_norm": 0.7518478631973267, "learning_rate": 9.197709951682268e-05, "loss": 0.012909440323710442, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.01299, "step": 120, "tokens/total": 3627168, "tokens/train_per_sec_per_gpu": 35.58, "tokens/trainable": 54896} +{"epoch": 0.47265625, "grad_norm": 0.15531380474567413, "learning_rate": 9.181089418799428e-05, "loss": 0.0027201471384614706, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00272, "step": 121, "tokens/total": 3657632, "tokens/train_per_sec_per_gpu": 38.07, "tokens/trainable": 55379} +{"epoch": 0.4765625, "grad_norm": 0.7706344127655029, "learning_rate": 9.164315700760271e-05, "loss": 0.009921186603605747, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00997, "step": 122, "tokens/total": 3687824, "tokens/train_per_sec_per_gpu": 32.0, "tokens/trainable": 55803} +{"epoch": 0.48046875, "grad_norm": 0.32032445073127747, "learning_rate": 9.147389495587671e-05, "loss": 0.006977465003728867, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.007, "step": 123, "tokens/total": 3718320, "tokens/train_per_sec_per_gpu": 36.45, "tokens/trainable": 56249} +{"epoch": 0.484375, "grad_norm": 0.3677634298801422, "learning_rate": 9.130311507650116e-05, "loss": 0.009076269343495369, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00912, "step": 124, "tokens/total": 3748672, "tokens/train_per_sec_per_gpu": 32.41, "tokens/trainable": 56713} +{"epoch": 0.48828125, "grad_norm": 0.27315884828567505, "learning_rate": 9.113082447632394e-05, "loss": 0.007012368179857731, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00704, "step": 125, "tokens/total": 3778976, "tokens/train_per_sec_per_gpu": 39.4, "tokens/trainable": 57196} +{"epoch": 0.4921875, "grad_norm": 0.48127928376197815, "learning_rate": 9.09570303250602e-05, "loss": 0.01003308966755867, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01008, "step": 126, "tokens/total": 3809296, "tokens/train_per_sec_per_gpu": 36.81, "tokens/trainable": 57680} +{"epoch": 0.49609375, "grad_norm": 0.17162159085273743, "learning_rate": 9.078173985499394e-05, "loss": 0.003832863178104162, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00384, "step": 127, "tokens/total": 3839328, "tokens/train_per_sec_per_gpu": 32.13, "tokens/trainable": 58110} +{"epoch": 0.5, "grad_norm": 0.20630274713039398, "learning_rate": 9.060496036067713e-05, "loss": 0.005646268837153912, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00566, "step": 128, "tokens/total": 3869824, "tokens/train_per_sec_per_gpu": 31.98, "tokens/trainable": 58534} +{"epoch": 0.50390625, "grad_norm": 0.12280423939228058, "learning_rate": 9.042669919862615e-05, "loss": 0.004154721274971962, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00416, "step": 129, "tokens/total": 3900080, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 58998} +{"epoch": 0.5078125, "grad_norm": 0.5695314407348633, "learning_rate": 9.024696378701557e-05, "loss": 0.02694959193468094, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.02732, "step": 130, "tokens/total": 3930560, "tokens/train_per_sec_per_gpu": 32.03, "tokens/trainable": 59448} +{"epoch": 0.51171875, "grad_norm": 0.24732492864131927, "learning_rate": 9.006576160536948e-05, "loss": 0.006222270894795656, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00624, "step": 131, "tokens/total": 3960496, "tokens/train_per_sec_per_gpu": 32.29, "tokens/trainable": 59906} +{"epoch": 0.515625, "grad_norm": 0.5292224287986755, "learning_rate": 8.988310019425035e-05, "loss": 0.016629355028271675, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01677, "step": 132, "tokens/total": 3990928, "tokens/train_per_sec_per_gpu": 34.39, "tokens/trainable": 60350} +{"epoch": 0.51953125, "grad_norm": 0.2769949734210968, "learning_rate": 8.969898715494506e-05, "loss": 0.011856021359562874, "memory/device_reserved (GiB)": 37.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01193, "step": 133, "tokens/total": 4021264, "tokens/train_per_sec_per_gpu": 36.35, "tokens/trainable": 60831} +{"epoch": 0.5234375, "grad_norm": 0.17929014563560486, "learning_rate": 8.951343014914869e-05, "loss": 0.007177118211984634, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0072, "step": 134, "tokens/total": 4051728, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 61305} +{"epoch": 0.52734375, "grad_norm": 0.14254862070083618, "learning_rate": 8.932643689864568e-05, "loss": 0.004720945842564106, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00473, "step": 135, "tokens/total": 4081920, "tokens/train_per_sec_per_gpu": 37.9, "tokens/trainable": 61792} +{"epoch": 0.53125, "grad_norm": 0.5255881547927856, "learning_rate": 8.913801518498845e-05, "loss": 0.001996636390686035, "memory/device_reserved (GiB)": 38.55, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.002, "step": 136, "tokens/total": 4112304, "tokens/train_per_sec_per_gpu": 31.59, "tokens/trainable": 62253} +{"epoch": 0.53515625, "grad_norm": 0.1990693211555481, "learning_rate": 8.894817284917364e-05, "loss": 0.007640001829713583, "memory/device_reserved (GiB)": 38.55, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00767, "step": 137, "tokens/total": 4142512, "tokens/train_per_sec_per_gpu": 31.86, "tokens/trainable": 62707} +{"epoch": 0.5390625, "grad_norm": 0.22452563047409058, "learning_rate": 8.875691779131569e-05, "loss": 0.005564739461988211, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00558, "step": 138, "tokens/total": 4172992, "tokens/train_per_sec_per_gpu": 29.52, "tokens/trainable": 63145} +{"epoch": 0.54296875, "grad_norm": 0.04953262209892273, "learning_rate": 8.856425797031829e-05, "loss": 0.0010842813644558191, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00108, "step": 139, "tokens/total": 4203136, "tokens/train_per_sec_per_gpu": 34.22, "tokens/trainable": 63587} +{"epoch": 0.546875, "grad_norm": 0.2684762179851532, "learning_rate": 8.837020140354295e-05, "loss": 0.006747591774910688, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00677, "step": 140, "tokens/total": 4233456, "tokens/train_per_sec_per_gpu": 36.99, "tokens/trainable": 64052} +{"epoch": 0.55078125, "grad_norm": 0.29308217763900757, "learning_rate": 8.817475616647554e-05, "loss": 0.010469472967088223, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01052, "step": 141, "tokens/total": 4263728, "tokens/train_per_sec_per_gpu": 34.57, "tokens/trainable": 64526} +{"epoch": 0.5546875, "grad_norm": 0.12753430008888245, "learning_rate": 8.797793039239017e-05, "loss": 0.003552365582436323, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.00356, "step": 142, "tokens/total": 4294432, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 64983} +{"epoch": 0.55859375, "grad_norm": 0.2863495647907257, "learning_rate": 8.777973227201069e-05, "loss": 0.009831942617893219, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00988, "step": 143, "tokens/total": 4324960, "tokens/train_per_sec_per_gpu": 37.74, "tokens/trainable": 65492} +{"epoch": 0.5625, "grad_norm": 0.17299926280975342, "learning_rate": 8.758017005316988e-05, "loss": 0.0038146874867379665, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00382, "step": 144, "tokens/total": 4355424, "tokens/train_per_sec_per_gpu": 35.77, "tokens/trainable": 65925} +{"epoch": 0.56640625, "grad_norm": 0.20330418646335602, "learning_rate": 8.737925204046629e-05, "loss": 0.00698842853307724, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00701, "step": 145, "tokens/total": 4385712, "tokens/train_per_sec_per_gpu": 31.38, "tokens/trainable": 66383} +{"epoch": 0.5703125, "grad_norm": 0.2837633788585663, "learning_rate": 8.717698659491851e-05, "loss": 0.005667341873049736, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00568, "step": 146, "tokens/total": 4416016, "tokens/train_per_sec_per_gpu": 31.4, "tokens/trainable": 66840} +{"epoch": 0.57421875, "grad_norm": 0.4520537257194519, "learning_rate": 8.697338213361735e-05, "loss": 0.015148473903536797, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01526, "step": 147, "tokens/total": 4446368, "tokens/train_per_sec_per_gpu": 31.72, "tokens/trainable": 67297} +{"epoch": 0.578125, "grad_norm": 0.3790777623653412, "learning_rate": 8.676844712937552e-05, "loss": 0.00472446670755744, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00474, "step": 148, "tokens/total": 4476880, "tokens/train_per_sec_per_gpu": 37.34, "tokens/trainable": 67769} +{"epoch": 0.58203125, "grad_norm": 0.7419846653938293, "learning_rate": 8.656219011037509e-05, "loss": 0.014075051993131638, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01417, "step": 149, "tokens/total": 4507232, "tokens/train_per_sec_per_gpu": 32.82, "tokens/trainable": 68257} +{"epoch": 0.5859375, "grad_norm": 0.26024505496025085, "learning_rate": 8.63546196598125e-05, "loss": 0.00536840595304966, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00538, "step": 150, "tokens/total": 4537376, "tokens/train_per_sec_per_gpu": 35.48, "tokens/trainable": 68706} +{"epoch": 0.58984375, "grad_norm": 0.41573861241340637, "learning_rate": 8.614574441554145e-05, "loss": 0.012942081317305565, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01303, "step": 151, "tokens/total": 4567584, "tokens/train_per_sec_per_gpu": 33.33, "tokens/trainable": 69131} +{"epoch": 0.59375, "grad_norm": 0.4435819685459137, "learning_rate": 8.593557306971349e-05, "loss": 0.010843953117728233, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0109, "step": 152, "tokens/total": 4597792, "tokens/train_per_sec_per_gpu": 31.93, "tokens/trainable": 69586} +{"epoch": 0.59765625, "grad_norm": 0.127182275056839, "learning_rate": 8.572411436841618e-05, "loss": 0.002272919053211808, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00228, "step": 153, "tokens/total": 4627936, "tokens/train_per_sec_per_gpu": 32.75, "tokens/trainable": 70061} +{"epoch": 0.6015625, "grad_norm": 0.08462858945131302, "learning_rate": 8.551137711130922e-05, "loss": 0.0021062777377665043, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00211, "step": 154, "tokens/total": 4658352, "tokens/train_per_sec_per_gpu": 27.9, "tokens/trainable": 70467} +{"epoch": 0.60546875, "grad_norm": 0.19869227707386017, "learning_rate": 8.529737015125824e-05, "loss": 0.005645486060529947, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00566, "step": 155, "tokens/total": 4688896, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 70933} +{"epoch": 0.609375, "grad_norm": 0.11944875866174698, "learning_rate": 8.508210239396639e-05, "loss": 0.0027996893040835857, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0028, "step": 156, "tokens/total": 4719168, "tokens/train_per_sec_per_gpu": 34.27, "tokens/trainable": 71382} +{"epoch": 0.61328125, "grad_norm": 0.07197802513837814, "learning_rate": 8.486558279760375e-05, "loss": 0.0010572505416348577, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00106, "step": 157, "tokens/total": 4749136, "tokens/train_per_sec_per_gpu": 30.75, "tokens/trainable": 71808} +{"epoch": 0.6171875, "grad_norm": 0.21313846111297607, "learning_rate": 8.464782037243449e-05, "loss": 0.012330076657235622, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.01241, "step": 158, "tokens/total": 4779472, "tokens/train_per_sec_per_gpu": 32.75, "tokens/trainable": 72249} +{"epoch": 0.62109375, "grad_norm": 0.2562062740325928, "learning_rate": 8.442882418044202e-05, "loss": 0.010313848033547401, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.01037, "step": 159, "tokens/total": 4809632, "tokens/train_per_sec_per_gpu": 35.1, "tokens/trainable": 72726} +{"epoch": 0.625, "grad_norm": 0.2375066578388214, "learning_rate": 8.420860333495179e-05, "loss": 0.01403950247913599, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.01414, "step": 160, "tokens/total": 4840112, "tokens/train_per_sec_per_gpu": 32.37, "tokens/trainable": 73148} +{"epoch": 0.62890625, "grad_norm": 0.18166372179985046, "learning_rate": 8.398716700025208e-05, "loss": 0.011304542422294617, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 34.04, "memory/max_allocated (GiB)": 34.04, "ppl": 1.01137, "step": 161, "tokens/total": 4870656, "tokens/train_per_sec_per_gpu": 31.42, "tokens/trainable": 73600} +{"epoch": 0.6328125, "grad_norm": 0.2600754499435425, "learning_rate": 8.376452439121266e-05, "loss": 0.012219483032822609, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.01229, "step": 162, "tokens/total": 4900608, "tokens/train_per_sec_per_gpu": 35.87, "tokens/trainable": 74068} +{"epoch": 0.63671875, "grad_norm": 0.11763953417539597, "learning_rate": 8.354068477290124e-05, "loss": 0.0028051501139998436, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00281, "step": 163, "tokens/total": 4930752, "tokens/train_per_sec_per_gpu": 35.92, "tokens/trainable": 74527} +{"epoch": 0.640625, "grad_norm": 0.166421040892601, "learning_rate": 8.331565746019807e-05, "loss": 0.007962819188833237, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00799, "step": 164, "tokens/total": 4961008, "tokens/train_per_sec_per_gpu": 32.01, "tokens/trainable": 74992} +{"epoch": 0.64453125, "grad_norm": 0.21651366353034973, "learning_rate": 8.308945181740812e-05, "loss": 0.008964363485574722, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.009, "step": 165, "tokens/total": 4991200, "tokens/train_per_sec_per_gpu": 35.46, "tokens/trainable": 75442} +{"epoch": 0.6484375, "grad_norm": 0.22853818535804749, "learning_rate": 8.286207725787153e-05, "loss": 0.008626708760857582, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00866, "step": 166, "tokens/total": 5021600, "tokens/train_per_sec_per_gpu": 31.44, "tokens/trainable": 75887} +{"epoch": 0.65234375, "grad_norm": 0.623465895652771, "learning_rate": 8.263354324357182e-05, "loss": 0.01683780550956726, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.01698, "step": 167, "tokens/total": 5052032, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 76331} +{"epoch": 0.65625, "grad_norm": 0.11736541986465454, "learning_rate": 8.240385928474219e-05, "loss": 0.004859982058405876, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00487, "step": 168, "tokens/total": 5080256, "tokens/train_per_sec_per_gpu": 36.11, "tokens/trainable": 76745} +{"epoch": 0.66015625, "grad_norm": 0.09870114177465439, "learning_rate": 8.217303493946967e-05, "loss": 0.0010998334037140012, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.0011, "step": 169, "tokens/total": 5110784, "tokens/train_per_sec_per_gpu": 34.47, "tokens/trainable": 77201} +{"epoch": 0.6640625, "grad_norm": 0.26467517018318176, "learning_rate": 8.194107981329746e-05, "loss": 0.010279067791998386, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.01033, "step": 170, "tokens/total": 5141200, "tokens/train_per_sec_per_gpu": 33.57, "tokens/trainable": 77655} +{"epoch": 0.66796875, "grad_norm": 0.02236294001340866, "learning_rate": 8.170800355882518e-05, "loss": 0.00039101624861359596, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00039, "step": 171, "tokens/total": 5171760, "tokens/train_per_sec_per_gpu": 39.4, "tokens/trainable": 78122} +{"epoch": 0.671875, "grad_norm": 0.1258402019739151, "learning_rate": 8.147381587530713e-05, "loss": 0.002418526913970709, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00242, "step": 172, "tokens/total": 5202272, "tokens/train_per_sec_per_gpu": 33.54, "tokens/trainable": 78576} +{"epoch": 0.67578125, "grad_norm": 0.20563526451587677, "learning_rate": 8.123852650824877e-05, "loss": 0.00472432654350996, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00474, "step": 173, "tokens/total": 5232800, "tokens/train_per_sec_per_gpu": 36.44, "tokens/trainable": 79059} +{"epoch": 0.6796875, "grad_norm": 0.5146963000297546, "learning_rate": 8.100214524900103e-05, "loss": 0.0075483620166778564, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.00758, "step": 174, "tokens/total": 5262672, "tokens/train_per_sec_per_gpu": 29.64, "tokens/trainable": 79498} +{"epoch": 0.68359375, "grad_norm": 0.124300517141819, "learning_rate": 8.076468193435301e-05, "loss": 0.002081199549138546, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00208, "step": 175, "tokens/total": 5293072, "tokens/train_per_sec_per_gpu": 30.34, "tokens/trainable": 79922} +{"epoch": 0.6875, "grad_norm": 0.28646209836006165, "learning_rate": 8.052614644612253e-05, "loss": 0.005611936096101999, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.46, "memory/max_allocated (GiB)": 33.46, "ppl": 1.00563, "step": 176, "tokens/total": 5321520, "tokens/train_per_sec_per_gpu": 35.56, "tokens/trainable": 80383} +{"epoch": 0.69140625, "grad_norm": 0.10610288381576538, "learning_rate": 8.028654871074489e-05, "loss": 0.0017271674005314708, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00173, "step": 177, "tokens/total": 5351904, "tokens/train_per_sec_per_gpu": 33.02, "tokens/trainable": 80824} +{"epoch": 0.6953125, "grad_norm": 0.04259132966399193, "learning_rate": 8.004589869885986e-05, "loss": 0.0009302249527536333, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00093, "step": 178, "tokens/total": 5382432, "tokens/train_per_sec_per_gpu": 32.21, "tokens/trainable": 81243} +{"epoch": 0.69921875, "grad_norm": 0.28697657585144043, "learning_rate": 7.980420642489674e-05, "loss": 0.003392249345779419, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0034, "step": 179, "tokens/total": 5412960, "tokens/train_per_sec_per_gpu": 34.27, "tokens/trainable": 81716} +{"epoch": 0.703125, "grad_norm": 0.4949120581150055, "learning_rate": 7.95614819466576e-05, "loss": 0.006915673613548279, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00694, "step": 180, "tokens/total": 5443232, "tokens/train_per_sec_per_gpu": 35.57, "tokens/trainable": 82181} +{"epoch": 0.70703125, "grad_norm": 0.8145425319671631, "learning_rate": 7.931773536489872e-05, "loss": 0.0152358990162611, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.29, "memory/max_allocated (GiB)": 33.29, "ppl": 1.01535, "step": 181, "tokens/total": 5471440, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 82626} +{"epoch": 0.7109375, "grad_norm": 0.2522812783718109, "learning_rate": 7.907297682291035e-05, "loss": 0.002765808254480362, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00277, "step": 182, "tokens/total": 5501744, "tokens/train_per_sec_per_gpu": 35.18, "tokens/trainable": 83089} +{"epoch": 0.71484375, "grad_norm": 0.017335012555122375, "learning_rate": 7.882721650609442e-05, "loss": 0.00021531448874156922, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00022, "step": 183, "tokens/total": 5532272, "tokens/train_per_sec_per_gpu": 35.53, "tokens/trainable": 83590} +{"epoch": 0.71875, "grad_norm": 0.6241586208343506, "learning_rate": 7.85804646415409e-05, "loss": 0.011036691255867481, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0111, "step": 184, "tokens/total": 5562784, "tokens/train_per_sec_per_gpu": 35.49, "tokens/trainable": 84046} +{"epoch": 0.72265625, "grad_norm": 0.012950743548572063, "learning_rate": 7.833273149760207e-05, "loss": 0.0002598769497126341, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00026, "step": 185, "tokens/total": 5592800, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 84517} +{"epoch": 0.7265625, "grad_norm": 1.1277121305465698, "learning_rate": 7.808402738346527e-05, "loss": 0.030348284170031548, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.03081, "step": 186, "tokens/total": 5623088, "tokens/train_per_sec_per_gpu": 34.04, "tokens/trainable": 84974} +{"epoch": 0.73046875, "grad_norm": 0.05753614753484726, "learning_rate": 7.783436264872382e-05, "loss": 0.0009072509710676968, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00091, "step": 187, "tokens/total": 5653344, "tokens/train_per_sec_per_gpu": 35.12, "tokens/trainable": 85460} +{"epoch": 0.734375, "grad_norm": 0.09686044603586197, "learning_rate": 7.758374768294647e-05, "loss": 0.00216390541754663, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00217, "step": 188, "tokens/total": 5683600, "tokens/train_per_sec_per_gpu": 36.08, "tokens/trainable": 85939} +{"epoch": 0.73828125, "grad_norm": 0.19125881791114807, "learning_rate": 7.733219291524489e-05, "loss": 0.005449399817734957, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00546, "step": 189, "tokens/total": 5711952, "tokens/train_per_sec_per_gpu": 38.69, "tokens/trainable": 86377} +{"epoch": 0.7421875, "grad_norm": 0.12624679505825043, "learning_rate": 7.707970881383977e-05, "loss": 0.0032629563938826323, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00327, "step": 190, "tokens/total": 5742336, "tokens/train_per_sec_per_gpu": 34.14, "tokens/trainable": 86834} +{"epoch": 0.74609375, "grad_norm": 0.18430523574352264, "learning_rate": 7.682630588562518e-05, "loss": 0.005647260695695877, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00566, "step": 191, "tokens/total": 5772560, "tokens/train_per_sec_per_gpu": 32.31, "tokens/trainable": 87265} +{"epoch": 0.75, "grad_norm": 0.12561790645122528, "learning_rate": 7.657199467573129e-05, "loss": 0.004560114815831184, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00457, "step": 192, "tokens/total": 5803136, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 87747} +{"epoch": 0.75390625, "grad_norm": 0.18357737362384796, "learning_rate": 7.631678576708561e-05, "loss": 0.005350396502763033, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00536, "step": 193, "tokens/total": 5833440, "tokens/train_per_sec_per_gpu": 36.45, "tokens/trainable": 88239} +{"epoch": 0.7578125, "grad_norm": 0.06960881501436234, "learning_rate": 7.606068977997255e-05, "loss": 0.0013192538172006607, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00132, "step": 194, "tokens/total": 5863552, "tokens/train_per_sec_per_gpu": 36.52, "tokens/trainable": 88718} +{"epoch": 0.76171875, "grad_norm": 0.41343674063682556, "learning_rate": 7.580371737159148e-05, "loss": 0.016355883330106735, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01649, "step": 195, "tokens/total": 5893680, "tokens/train_per_sec_per_gpu": 31.46, "tokens/trainable": 89129} +{"epoch": 0.765625, "grad_norm": 0.05512367933988571, "learning_rate": 7.554587923561324e-05, "loss": 0.0009896137053146958, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00099, "step": 196, "tokens/total": 5923744, "tokens/train_per_sec_per_gpu": 32.48, "tokens/trainable": 89567} +{"epoch": 0.76953125, "grad_norm": 0.3041112422943115, "learning_rate": 7.528718610173511e-05, "loss": 0.008880488574504852, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00892, "step": 197, "tokens/total": 5954128, "tokens/train_per_sec_per_gpu": 30.33, "tokens/trainable": 89988} +{"epoch": 0.7734375, "grad_norm": 0.035138241946697235, "learning_rate": 7.502764873523431e-05, "loss": 0.0007115794578567147, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00071, "step": 198, "tokens/total": 5984352, "tokens/train_per_sec_per_gpu": 36.27, "tokens/trainable": 90434} +{"epoch": 0.77734375, "grad_norm": 0.17300349473953247, "learning_rate": 7.476727793652011e-05, "loss": 0.0062064481899142265, "memory/device_reserved (GiB)": 35.6, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00623, "step": 199, "tokens/total": 6014704, "tokens/train_per_sec_per_gpu": 34.38, "tokens/trainable": 90913} +{"epoch": 0.78125, "grad_norm": 0.11639569699764252, "learning_rate": 7.450608454068415e-05, "loss": 0.0035372006241232157, "memory/device_reserved (GiB)": 35.66, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00354, "step": 200, "tokens/total": 6045056, "tokens/train_per_sec_per_gpu": 34.44, "tokens/trainable": 91355} +{"epoch": 0.78515625, "grad_norm": 0.3703908920288086, "learning_rate": 7.424407941704987e-05, "loss": 0.00820743478834629, "memory/device_reserved (GiB)": 35.66, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00824, "step": 201, "tokens/total": 6075504, "tokens/train_per_sec_per_gpu": 37.51, "tokens/trainable": 91856} +{"epoch": 0.7890625, "grad_norm": 0.2921426594257355, "learning_rate": 7.398127346871986e-05, "loss": 0.007985620759427547, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00802, "step": 202, "tokens/total": 6105904, "tokens/train_per_sec_per_gpu": 34.36, "tokens/trainable": 92311} +{"epoch": 0.79296875, "grad_norm": 0.059785567224025726, "learning_rate": 7.371767763212238e-05, "loss": 0.0009134798310697079, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00091, "step": 203, "tokens/total": 6136272, "tokens/train_per_sec_per_gpu": 34.96, "tokens/trainable": 92764} +{"epoch": 0.796875, "grad_norm": 0.1626388430595398, "learning_rate": 7.345330287655617e-05, "loss": 0.0032109033782035112, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.65, "memory/max_allocated (GiB)": 33.65, "ppl": 1.00322, "step": 204, "tokens/total": 6166336, "tokens/train_per_sec_per_gpu": 33.24, "tokens/trainable": 93227} +{"epoch": 0.80078125, "grad_norm": 0.3314966857433319, "learning_rate": 7.31881602037339e-05, "loss": 0.0066377827897667885, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00666, "step": 205, "tokens/total": 6196720, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 93675} +{"epoch": 0.8046875, "grad_norm": 0.13114461302757263, "learning_rate": 7.29222606473245e-05, "loss": 0.003180722938850522, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00319, "step": 206, "tokens/total": 6227424, "tokens/train_per_sec_per_gpu": 32.38, "tokens/trainable": 94120} +{"epoch": 0.80859375, "grad_norm": 0.014965695329010487, "learning_rate": 7.265561527249383e-05, "loss": 0.00022718981199432164, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00023, "step": 207, "tokens/total": 6257616, "tokens/train_per_sec_per_gpu": 35.24, "tokens/trainable": 94557} +{"epoch": 0.8125, "grad_norm": 0.04264510050415993, "learning_rate": 7.238823517544436e-05, "loss": 0.000695791095495224, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0007, "step": 208, "tokens/total": 6288000, "tokens/train_per_sec_per_gpu": 40.72, "tokens/trainable": 95035} +{"epoch": 0.81640625, "grad_norm": 0.054736051708459854, "learning_rate": 7.212013148295333e-05, "loss": 0.0007617401424795389, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00076, "step": 209, "tokens/total": 6318336, "tokens/train_per_sec_per_gpu": 37.65, "tokens/trainable": 95517} +{"epoch": 0.8203125, "grad_norm": 0.1429799348115921, "learning_rate": 7.185131535190975e-05, "loss": 0.003273893380537629, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00328, "step": 210, "tokens/total": 6348656, "tokens/train_per_sec_per_gpu": 36.59, "tokens/trainable": 96026} +{"epoch": 0.82421875, "grad_norm": 0.00420429976657033, "learning_rate": 7.158179796885005e-05, "loss": 5.528773181140423e-05, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00006, "step": 211, "tokens/total": 6379040, "tokens/train_per_sec_per_gpu": 35.72, "tokens/trainable": 96477} +{"epoch": 0.828125, "grad_norm": 0.0016959067434072495, "learning_rate": 7.131159054949273e-05, "loss": 3.3626260119490325e-05, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00003, "step": 212, "tokens/total": 6409312, "tokens/train_per_sec_per_gpu": 33.24, "tokens/trainable": 96951} +{"epoch": 0.83203125, "grad_norm": 0.0643712654709816, "learning_rate": 7.104070433827139e-05, "loss": 0.00035337061854079366, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00035, "step": 213, "tokens/total": 6439520, "tokens/train_per_sec_per_gpu": 30.95, "tokens/trainable": 97350} +{"epoch": 0.8359375, "grad_norm": 0.001674463041126728, "learning_rate": 7.076915060786705e-05, "loss": 1.70804687513737e-05, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 214, "tokens/total": 6469888, "tokens/train_per_sec_per_gpu": 29.99, "tokens/trainable": 97792} +{"epoch": 0.83984375, "grad_norm": 0.20266881585121155, "learning_rate": 7.049694065873882e-05, "loss": 0.0047647226601839066, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00478, "step": 215, "tokens/total": 6500128, "tokens/train_per_sec_per_gpu": 36.56, "tokens/trainable": 98257} +{"epoch": 0.84375, "grad_norm": 0.03204328939318657, "learning_rate": 7.022408581865382e-05, "loss": 0.0003669401630759239, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00037, "step": 216, "tokens/total": 6530832, "tokens/train_per_sec_per_gpu": 32.71, "tokens/trainable": 98731} +{"epoch": 0.84765625, "grad_norm": 0.006207747384905815, "learning_rate": 6.99505974422157e-05, "loss": 8.174816321115941e-05, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00008, "step": 217, "tokens/total": 6561248, "tokens/train_per_sec_per_gpu": 36.55, "tokens/trainable": 99212} +{"epoch": 0.8515625, "grad_norm": 0.2884034216403961, "learning_rate": 6.967648691039213e-05, "loss": 0.0017384591046720743, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00174, "step": 218, "tokens/total": 6591648, "tokens/train_per_sec_per_gpu": 33.2, "tokens/trainable": 99654} +{"epoch": 0.85546875, "grad_norm": 0.0034673702903091908, "learning_rate": 6.940176563004123e-05, "loss": 3.302653203718364e-05, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00003, "step": 219, "tokens/total": 6622368, "tokens/train_per_sec_per_gpu": 33.02, "tokens/trainable": 100086} +{"epoch": 0.859375, "grad_norm": 0.15575405955314636, "learning_rate": 6.912644503343682e-05, "loss": 0.0013022763887420297, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0013, "step": 220, "tokens/total": 6652848, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 100524} +{"epoch": 0.86328125, "grad_norm": 0.0012149910908192396, "learning_rate": 6.885053657779273e-05, "loss": 1.713100755296182e-05, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00002, "step": 221, "tokens/total": 6683392, "tokens/train_per_sec_per_gpu": 33.53, "tokens/trainable": 100996} +{"epoch": 0.8671875, "grad_norm": 0.031321559101343155, "learning_rate": 6.857405174478604e-05, "loss": 0.00015568418893963099, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00016, "step": 222, "tokens/total": 6713712, "tokens/train_per_sec_per_gpu": 32.1, "tokens/trainable": 101449} +{"epoch": 0.87109375, "grad_norm": 0.5521696209907532, "learning_rate": 6.82970020400792e-05, "loss": 0.00847858190536499, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00851, "step": 223, "tokens/total": 6744176, "tokens/train_per_sec_per_gpu": 33.2, "tokens/trainable": 101905} +{"epoch": 0.875, "grad_norm": 0.5790396928787231, "learning_rate": 6.801939899284132e-05, "loss": 0.013318291865289211, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01341, "step": 224, "tokens/total": 6774416, "tokens/train_per_sec_per_gpu": 36.67, "tokens/trainable": 102411} +{"epoch": 0.87890625, "grad_norm": 0.6164970993995667, "learning_rate": 6.774125415526827e-05, "loss": 0.013708921149373055, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.0138, "step": 225, "tokens/total": 6804592, "tokens/train_per_sec_per_gpu": 34.02, "tokens/trainable": 102882} +{"epoch": 0.8828125, "grad_norm": 0.009372257627546787, "learning_rate": 6.746257910210214e-05, "loss": 0.00011854815966216847, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00012, "step": 226, "tokens/total": 6834976, "tokens/train_per_sec_per_gpu": 34.73, "tokens/trainable": 103332} +{"epoch": 0.88671875, "grad_norm": 0.02493397518992424, "learning_rate": 6.718338543014937e-05, "loss": 0.00019317460828460753, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00019, "step": 227, "tokens/total": 6865408, "tokens/train_per_sec_per_gpu": 35.3, "tokens/trainable": 103790} +{"epoch": 0.890625, "grad_norm": 0.007950029335916042, "learning_rate": 6.69036847577983e-05, "loss": 6.161297642393038e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00006, "step": 228, "tokens/total": 6895664, "tokens/train_per_sec_per_gpu": 34.51, "tokens/trainable": 104246} +{"epoch": 0.89453125, "grad_norm": 0.9438257813453674, "learning_rate": 6.662348872453553e-05, "loss": 0.01598445326089859, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01611, "step": 229, "tokens/total": 6926016, "tokens/train_per_sec_per_gpu": 37.87, "tokens/trainable": 104707} +{"epoch": 0.8984375, "grad_norm": 0.04287649691104889, "learning_rate": 6.63428089904618e-05, "loss": 0.0005880278768017888, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00059, "step": 230, "tokens/total": 6956384, "tokens/train_per_sec_per_gpu": 37.02, "tokens/trainable": 105167} +{"epoch": 0.90234375, "grad_norm": 0.275064617395401, "learning_rate": 6.60616572358065e-05, "loss": 0.002924790605902672, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00293, "step": 231, "tokens/total": 6986768, "tokens/train_per_sec_per_gpu": 33.3, "tokens/trainable": 105576} +{"epoch": 0.90625, "grad_norm": 0.19834892451763153, "learning_rate": 6.578004516044172e-05, "loss": 0.003120731096714735, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00313, "step": 232, "tokens/total": 7017232, "tokens/train_per_sec_per_gpu": 37.01, "tokens/trainable": 106043} +{"epoch": 0.91015625, "grad_norm": 0.031313009560108185, "learning_rate": 6.549798448339548e-05, "loss": 0.0006567268865182996, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00066, "step": 233, "tokens/total": 7047568, "tokens/train_per_sec_per_gpu": 35.14, "tokens/trainable": 106506} +{"epoch": 0.9140625, "grad_norm": 0.029670486226677895, "learning_rate": 6.521548694236384e-05, "loss": 0.00038576917722821236, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00039, "step": 234, "tokens/total": 7077760, "tokens/train_per_sec_per_gpu": 34.22, "tokens/trainable": 106951} +{"epoch": 0.91796875, "grad_norm": 0.13840647041797638, "learning_rate": 6.493256429322259e-05, "loss": 0.003701703157275915, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00371, "step": 235, "tokens/total": 7107760, "tokens/train_per_sec_per_gpu": 30.87, "tokens/trainable": 107382} +{"epoch": 0.921875, "grad_norm": 0.19832508265972137, "learning_rate": 6.464922830953799e-05, "loss": 0.008310232311487198, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00834, "step": 236, "tokens/total": 7138144, "tokens/train_per_sec_per_gpu": 33.03, "tokens/trainable": 107814} +{"epoch": 0.92578125, "grad_norm": 0.3925366699695587, "learning_rate": 6.436549078207688e-05, "loss": 0.006797720678150654, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00682, "step": 237, "tokens/total": 7168352, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 108274} +{"epoch": 0.9296875, "grad_norm": 0.029645387083292007, "learning_rate": 6.408136351831592e-05, "loss": 0.0006107640801928937, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00061, "step": 238, "tokens/total": 7198624, "tokens/train_per_sec_per_gpu": 29.31, "tokens/trainable": 108714} +{"epoch": 0.93359375, "grad_norm": 0.06903097778558731, "learning_rate": 6.379685834195036e-05, "loss": 0.0018063001334667206, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00181, "step": 239, "tokens/total": 7229216, "tokens/train_per_sec_per_gpu": 36.89, "tokens/trainable": 109220} +{"epoch": 0.9375, "grad_norm": 0.027704427018761635, "learning_rate": 6.351198709240186e-05, "loss": 0.0005392928142100573, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00054, "step": 240, "tokens/total": 7259648, "tokens/train_per_sec_per_gpu": 32.2, "tokens/trainable": 109672} +{"epoch": 0.94140625, "grad_norm": 0.17064078152179718, "learning_rate": 6.32267616243259e-05, "loss": 0.002919984282925725, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00292, "step": 241, "tokens/total": 7289824, "tokens/train_per_sec_per_gpu": 35.3, "tokens/trainable": 110135} +{"epoch": 0.9453125, "grad_norm": 0.16251814365386963, "learning_rate": 6.294119380711849e-05, "loss": 0.005423969589173794, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00544, "step": 242, "tokens/total": 7320288, "tokens/train_per_sec_per_gpu": 30.68, "tokens/trainable": 110563} +{"epoch": 0.94921875, "grad_norm": 0.20398780703544617, "learning_rate": 6.265529552442209e-05, "loss": 0.0033713094890117645, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00338, "step": 243, "tokens/total": 7350736, "tokens/train_per_sec_per_gpu": 37.21, "tokens/trainable": 111049} +{"epoch": 0.953125, "grad_norm": 0.06107059493660927, "learning_rate": 6.236907867363127e-05, "loss": 0.001296353293582797, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.0013, "step": 244, "tokens/total": 7381280, "tokens/train_per_sec_per_gpu": 33.51, "tokens/trainable": 111495} +{"epoch": 0.95703125, "grad_norm": 0.07008613646030426, "learning_rate": 6.208255516539749e-05, "loss": 0.0011632217792794108, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00116, "step": 245, "tokens/total": 7411632, "tokens/train_per_sec_per_gpu": 31.94, "tokens/trainable": 111968} +{"epoch": 0.9609375, "grad_norm": 0.04514322429895401, "learning_rate": 6.179573692313344e-05, "loss": 0.0008365803514607251, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00084, "step": 246, "tokens/total": 7441904, "tokens/train_per_sec_per_gpu": 36.7, "tokens/trainable": 112416} +{"epoch": 0.96484375, "grad_norm": 0.11721615493297577, "learning_rate": 6.150863588251694e-05, "loss": 0.0016282327705994248, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00163, "step": 247, "tokens/total": 7472368, "tokens/train_per_sec_per_gpu": 34.16, "tokens/trainable": 112882} +{"epoch": 0.96875, "grad_norm": 0.04209491237998009, "learning_rate": 6.122126399099419e-05, "loss": 0.0007072857697494328, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00071, "step": 248, "tokens/total": 7502656, "tokens/train_per_sec_per_gpu": 40.32, "tokens/trainable": 113390} +{"epoch": 0.97265625, "grad_norm": 0.339526504278183, "learning_rate": 6.0933633207282615e-05, "loss": 0.008826559409499168, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00887, "step": 249, "tokens/total": 7532800, "tokens/train_per_sec_per_gpu": 37.92, "tokens/trainable": 113904} +{"epoch": 0.9765625, "grad_norm": 0.04211124777793884, "learning_rate": 6.064575550087316e-05, "loss": 0.0006392797804437578, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00064, "step": 250, "tokens/total": 7563264, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 114351} +{"epoch": 0.98046875, "grad_norm": 0.12704773247241974, "learning_rate": 6.0357642851532245e-05, "loss": 0.0020626899786293507, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00206, "step": 251, "tokens/total": 7593840, "tokens/train_per_sec_per_gpu": 35.32, "tokens/trainable": 114842} +{"epoch": 0.984375, "grad_norm": 0.05915632098913193, "learning_rate": 6.0069307248803294e-05, "loss": 0.0007249495247378945, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00073, "step": 252, "tokens/total": 7624416, "tokens/train_per_sec_per_gpu": 29.06, "tokens/trainable": 115263} +{"epoch": 0.98828125, "grad_norm": 0.04091682285070419, "learning_rate": 5.9780760691507635e-05, "loss": 0.0006673121242783964, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00067, "step": 253, "tokens/total": 7654688, "tokens/train_per_sec_per_gpu": 32.91, "tokens/trainable": 115703} +{"epoch": 0.9921875, "grad_norm": 0.03956187143921852, "learning_rate": 5.9492015187245334e-05, "loss": 0.0004259806009940803, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00043, "step": 254, "tokens/total": 7685088, "tokens/train_per_sec_per_gpu": 33.46, "tokens/trainable": 116157} +{"epoch": 0.99609375, "grad_norm": 0.0026254591066390276, "learning_rate": 5.920308275189541e-05, "loss": 5.904503632336855e-05, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00006, "step": 255, "tokens/total": 7715552, "tokens/train_per_sec_per_gpu": 32.15, "tokens/trainable": 116567} +{"epoch": 1.0, "grad_norm": 0.05956212058663368, "learning_rate": 5.8913975409115874e-05, "loss": 0.0007089228602126241, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00071, "step": 256, "tokens/total": 7745872, "tokens/train_per_sec_per_gpu": 31.4, "tokens/trainable": 117030} +{"epoch": 1.00390625, "grad_norm": 0.004725128877907991, "learning_rate": 5.8624705189843395e-05, "loss": 8.39156418805942e-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": 35.03, "tokens/trainable": 117517} +{"epoch": 1.0078125, "grad_norm": 0.0034643823746591806, "learning_rate": 5.833528413179249e-05, "loss": 6.966243381612003e-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": 38.58, "tokens/trainable": 118015} +{"epoch": 1.01171875, "grad_norm": 0.4294135570526123, "learning_rate": 5.80457242789548e-05, "loss": 0.004226258024573326, "memory/device_reserved (GiB)": 35.62, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00424, "step": 259, "tokens/total": 7836800, "tokens/train_per_sec_per_gpu": 36.26, "tokens/trainable": 118520} +{"epoch": 1.015625, "grad_norm": 0.10779248923063278, "learning_rate": 5.77560376810977e-05, "loss": 0.0016014521243050694, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0016, "step": 260, "tokens/total": 7867040, "tokens/train_per_sec_per_gpu": 36.37, "tokens/trainable": 118992} +{"epoch": 1.01953125, "grad_norm": 0.0030922323931008577, "learning_rate": 5.7466236393263005e-05, "loss": 4.83963631268125e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00005, "step": 261, "tokens/total": 7897408, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 119438} +{"epoch": 1.0234375, "grad_norm": 0.0015131091931834817, "learning_rate": 5.717633247526522e-05, "loss": 2.9313643608475104e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00003, "step": 262, "tokens/total": 7927648, "tokens/train_per_sec_per_gpu": 32.24, "tokens/trainable": 119881} +{"epoch": 1.02734375, "grad_norm": 0.023485446348786354, "learning_rate": 5.688633799118971e-05, "loss": 0.00011603974417084828, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00012, "step": 263, "tokens/total": 7957968, "tokens/train_per_sec_per_gpu": 28.01, "tokens/trainable": 120285} +{"epoch": 1.03125, "grad_norm": 0.007287763059139252, "learning_rate": 5.659626500889066e-05, "loss": 7.50380422687158e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.65, "memory/max_allocated (GiB)": 33.65, "ppl": 1.00008, "step": 264, "tokens/total": 7987888, "tokens/train_per_sec_per_gpu": 33.19, "tokens/trainable": 120755} +{"epoch": 1.03515625, "grad_norm": 0.019914504140615463, "learning_rate": 5.6306125599488905e-05, "loss": 0.0001366423675790429, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00014, "step": 265, "tokens/total": 8018048, "tokens/train_per_sec_per_gpu": 35.61, "tokens/trainable": 121194} +{"epoch": 1.0390625, "grad_norm": 0.03110283613204956, "learning_rate": 5.601593183686955e-05, "loss": 0.00017948952154256403, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00018, "step": 266, "tokens/total": 8048448, "tokens/train_per_sec_per_gpu": 36.75, "tokens/trainable": 121670} +{"epoch": 1.04296875, "grad_norm": 0.013412952423095703, "learning_rate": 5.572569579717961e-05, "loss": 0.00018778869707603008, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00019, "step": 267, "tokens/total": 8078848, "tokens/train_per_sec_per_gpu": 33.92, "tokens/trainable": 122142} +{"epoch": 1.046875, "grad_norm": 0.0006959482561796904, "learning_rate": 5.543542955832538e-05, "loss": 1.8885750250774436e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 268, "tokens/total": 8109360, "tokens/train_per_sec_per_gpu": 32.68, "tokens/trainable": 122585} +{"epoch": 1.05078125, "grad_norm": 0.15634001791477203, "learning_rate": 5.514514519946986e-05, "loss": 0.0009451522491872311, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00095, "step": 269, "tokens/total": 8139888, "tokens/train_per_sec_per_gpu": 39.13, "tokens/trainable": 123091} +{"epoch": 1.0546875, "grad_norm": 0.006704932544380426, "learning_rate": 5.485485480053015e-05, "loss": 8.153816452249885e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00008, "step": 270, "tokens/total": 8170480, "tokens/train_per_sec_per_gpu": 31.2, "tokens/trainable": 123505} +{"epoch": 1.05859375, "grad_norm": 0.04777071997523308, "learning_rate": 5.4564570441674645e-05, "loss": 0.00040322638233192265, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.41, "memory/max_allocated (GiB)": 33.41, "ppl": 1.0004, "step": 271, "tokens/total": 8198848, "tokens/train_per_sec_per_gpu": 35.71, "tokens/trainable": 123953} +{"epoch": 1.0625, "grad_norm": 0.28593015670776367, "learning_rate": 5.42743042028204e-05, "loss": 0.0016996364574879408, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0017, "step": 272, "tokens/total": 8229200, "tokens/train_per_sec_per_gpu": 28.82, "tokens/trainable": 124400} +{"epoch": 1.06640625, "grad_norm": 0.035160090774297714, "learning_rate": 5.3984068163130464e-05, "loss": 0.00030365801649168134, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0003, "step": 273, "tokens/total": 8259536, "tokens/train_per_sec_per_gpu": 35.12, "tokens/trainable": 124870} +{"epoch": 1.0703125, "grad_norm": 0.2729821503162384, "learning_rate": 5.369387440051111e-05, "loss": 0.0029124405700713396, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00292, "step": 274, "tokens/total": 8289904, "tokens/train_per_sec_per_gpu": 32.3, "tokens/trainable": 125333} +{"epoch": 1.07421875, "grad_norm": 0.0004543201648630202, "learning_rate": 5.340373499110935e-05, "loss": 1.239740231540054e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00001, "step": 275, "tokens/total": 8320176, "tokens/train_per_sec_per_gpu": 36.91, "tokens/trainable": 125834} +{"epoch": 1.078125, "grad_norm": 0.0025960030034184456, "learning_rate": 5.3113662008810304e-05, "loss": 3.4138491173507646e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00003, "step": 276, "tokens/total": 8350832, "tokens/train_per_sec_per_gpu": 38.3, "tokens/trainable": 126316} +{"epoch": 1.08203125, "grad_norm": 0.047371700406074524, "learning_rate": 5.282366752473479e-05, "loss": 0.0004324351029936224, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00043, "step": 277, "tokens/total": 8380976, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 126798} +{"epoch": 1.0859375, "grad_norm": 0.07045093923807144, "learning_rate": 5.2533763606737005e-05, "loss": 0.0006008927593939006, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.0006, "step": 278, "tokens/total": 8411072, "tokens/train_per_sec_per_gpu": 29.84, "tokens/trainable": 127223} +{"epoch": 1.08984375, "grad_norm": 0.00697721354663372, "learning_rate": 5.224396231890232e-05, "loss": 3.3106734917964786e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.00003, "step": 279, "tokens/total": 8440736, "tokens/train_per_sec_per_gpu": 36.49, "tokens/trainable": 127672} +{"epoch": 1.09375, "grad_norm": 0.6101890802383423, "learning_rate": 5.195427572104522e-05, "loss": 0.015062580816447735, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.01518, "step": 280, "tokens/total": 8470944, "tokens/train_per_sec_per_gpu": 32.38, "tokens/trainable": 128116} +{"epoch": 1.09765625, "grad_norm": 0.05356109142303467, "learning_rate": 5.166471586820751e-05, "loss": 0.0003783979918807745, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00038, "step": 281, "tokens/total": 8501104, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 128589} +{"epoch": 1.1015625, "grad_norm": 0.3603111505508423, "learning_rate": 5.1375294810156615e-05, "loss": 0.008839127607643604, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00888, "step": 282, "tokens/total": 8531696, "tokens/train_per_sec_per_gpu": 33.97, "tokens/trainable": 129036} +{"epoch": 1.10546875, "grad_norm": 0.000897557009011507, "learning_rate": 5.1086024590884144e-05, "loss": 1.7376631149090827e-05, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00002, "step": 283, "tokens/total": 8561936, "tokens/train_per_sec_per_gpu": 35.42, "tokens/trainable": 129485} +{"epoch": 1.109375, "grad_norm": 0.09921212494373322, "learning_rate": 5.079691724810461e-05, "loss": 0.001691737212240696, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00169, "step": 284, "tokens/total": 8592192, "tokens/train_per_sec_per_gpu": 29.77, "tokens/trainable": 129920} +{"epoch": 1.11328125, "grad_norm": 0.06643540412187576, "learning_rate": 5.0507984812754684e-05, "loss": 0.0005100784474052489, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00051, "step": 285, "tokens/total": 8622432, "tokens/train_per_sec_per_gpu": 36.35, "tokens/trainable": 130417} +{"epoch": 1.1171875, "grad_norm": 0.1579083353281021, "learning_rate": 5.021923930849237e-05, "loss": 0.0014161370927467942, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00142, "step": 286, "tokens/total": 8652768, "tokens/train_per_sec_per_gpu": 36.19, "tokens/trainable": 130903} +{"epoch": 1.12109375, "grad_norm": 0.060738325119018555, "learning_rate": 4.99306927511967e-05, "loss": 0.0004249056801199913, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00042, "step": 287, "tokens/total": 8683296, "tokens/train_per_sec_per_gpu": 35.79, "tokens/trainable": 131343} +{"epoch": 1.125, "grad_norm": 0.5256732702255249, "learning_rate": 4.964235714846775e-05, "loss": 0.011904106475412846, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01198, "step": 288, "tokens/total": 8713504, "tokens/train_per_sec_per_gpu": 32.36, "tokens/trainable": 131763} +{"epoch": 1.12890625, "grad_norm": 0.000379653851268813, "learning_rate": 4.9354244499126866e-05, "loss": 9.642227269068826e-06, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00001, "step": 289, "tokens/total": 8743888, "tokens/train_per_sec_per_gpu": 33.95, "tokens/trainable": 132229} +{"epoch": 1.1328125, "grad_norm": 0.0017015178455039859, "learning_rate": 4.90663667927174e-05, "loss": 1.8409156837151386e-05, "memory/device_reserved (GiB)": 35.54, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00002, "step": 290, "tokens/total": 8774336, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 132678} +{"epoch": 1.13671875, "grad_norm": 0.09125205129384995, "learning_rate": 4.877873600900581e-05, "loss": 0.0005971058853901923, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0006, "step": 291, "tokens/total": 8804816, "tokens/train_per_sec_per_gpu": 29.53, "tokens/trainable": 133136} +{"epoch": 1.140625, "grad_norm": 0.006331851705908775, "learning_rate": 4.849136411748306e-05, "loss": 6.921300519024953e-05, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00007, "step": 292, "tokens/total": 8835024, "tokens/train_per_sec_per_gpu": 35.52, "tokens/trainable": 133608} +{"epoch": 1.14453125, "grad_norm": 0.1789214313030243, "learning_rate": 4.8204263076866574e-05, "loss": 0.0013293421361595392, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00133, "step": 293, "tokens/total": 8865408, "tokens/train_per_sec_per_gpu": 40.63, "tokens/trainable": 134108} +{"epoch": 1.1484375, "grad_norm": 0.02855750173330307, "learning_rate": 4.791744483460251e-05, "loss": 0.00025921204360201955, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00026, "step": 294, "tokens/total": 8895824, "tokens/train_per_sec_per_gpu": 30.9, "tokens/trainable": 134541} +{"epoch": 1.15234375, "grad_norm": 0.02957119792699814, "learning_rate": 4.7630921326368736e-05, "loss": 0.00019251988851465285, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00019, "step": 295, "tokens/total": 8926336, "tokens/train_per_sec_per_gpu": 29.64, "tokens/trainable": 134966} +{"epoch": 1.15625, "grad_norm": 0.024680135771632195, "learning_rate": 4.7344704475577916e-05, "loss": 0.00031903525814414024, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00032, "step": 296, "tokens/total": 8956656, "tokens/train_per_sec_per_gpu": 33.73, "tokens/trainable": 135402} +{"epoch": 1.16015625, "grad_norm": 0.1391189694404602, "learning_rate": 4.705880619288153e-05, "loss": 0.0013685933081433177, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00137, "step": 297, "tokens/total": 8986752, "tokens/train_per_sec_per_gpu": 30.21, "tokens/trainable": 135804} +{"epoch": 1.1640625, "grad_norm": 0.0789211168885231, "learning_rate": 4.677323837567412e-05, "loss": 0.0005729947588406503, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00057, "step": 298, "tokens/total": 9017024, "tokens/train_per_sec_per_gpu": 31.3, "tokens/trainable": 136231} +{"epoch": 1.16796875, "grad_norm": 0.018962092697620392, "learning_rate": 4.6488012907598146e-05, "loss": 0.00019800761947408319, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0002, "step": 299, "tokens/total": 9047424, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 136705} +{"epoch": 1.171875, "grad_norm": 0.06257972121238708, "learning_rate": 4.620314165804964e-05, "loss": 0.00047805995563976467, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00048, "step": 300, "tokens/total": 9077648, "tokens/train_per_sec_per_gpu": 34.92, "tokens/trainable": 137178} +{"epoch": 1.17578125, "grad_norm": 0.13331402838230133, "learning_rate": 4.591863648168407e-05, "loss": 0.0005989698111079633, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.0006, "step": 301, "tokens/total": 9108032, "tokens/train_per_sec_per_gpu": 34.54, "tokens/trainable": 137643} +{"epoch": 1.1796875, "grad_norm": 0.049926936626434326, "learning_rate": 4.5634509217923135e-05, "loss": 0.0006213528104126453, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00062, "step": 302, "tokens/total": 9138240, "tokens/train_per_sec_per_gpu": 32.12, "tokens/trainable": 138078} +{"epoch": 1.18359375, "grad_norm": 0.6875918507575989, "learning_rate": 4.535077169046201e-05, "loss": 0.006850574631243944, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00687, "step": 303, "tokens/total": 9168352, "tokens/train_per_sec_per_gpu": 35.3, "tokens/trainable": 138515} +{"epoch": 1.1875, "grad_norm": 0.41269931197166443, "learning_rate": 4.506743570677743e-05, "loss": 0.008352026343345642, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00839, "step": 304, "tokens/total": 9198864, "tokens/train_per_sec_per_gpu": 35.35, "tokens/trainable": 138948} +{"epoch": 1.19140625, "grad_norm": 0.25145408511161804, "learning_rate": 4.478451305763618e-05, "loss": 0.004350180737674236, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00436, "step": 305, "tokens/total": 9229104, "tokens/train_per_sec_per_gpu": 38.28, "tokens/trainable": 139408} +{"epoch": 1.1953125, "grad_norm": 0.017037076875567436, "learning_rate": 4.450201551660454e-05, "loss": 0.0001216172895510681, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.38, "memory/max_allocated (GiB)": 33.38, "ppl": 1.00012, "step": 306, "tokens/total": 9257344, "tokens/train_per_sec_per_gpu": 33.61, "tokens/trainable": 139840} +{"epoch": 1.19921875, "grad_norm": 0.049017518758773804, "learning_rate": 4.4219954839558276e-05, "loss": 0.00043262774124741554, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00043, "step": 307, "tokens/total": 9287520, "tokens/train_per_sec_per_gpu": 33.1, "tokens/trainable": 140281} +{"epoch": 1.203125, "grad_norm": 0.020307866856455803, "learning_rate": 4.393834276419352e-05, "loss": 0.00011033139890059829, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00011, "step": 308, "tokens/total": 9317888, "tokens/train_per_sec_per_gpu": 37.6, "tokens/trainable": 140776} +{"epoch": 1.20703125, "grad_norm": 0.0014244886115193367, "learning_rate": 4.36571910095382e-05, "loss": 2.7976970159215853e-05, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00003, "step": 309, "tokens/total": 9348256, "tokens/train_per_sec_per_gpu": 36.96, "tokens/trainable": 141228} +{"epoch": 1.2109375, "grad_norm": 0.0005130386562086642, "learning_rate": 4.337651127546448e-05, "loss": 1.5001332030806225e-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": 37.59, "tokens/trainable": 141679} +{"epoch": 1.21484375, "grad_norm": 0.015154760330915451, "learning_rate": 4.3096315242201736e-05, "loss": 0.00014183670282363892, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00014, "step": 311, "tokens/total": 9408848, "tokens/train_per_sec_per_gpu": 32.51, "tokens/trainable": 142122} +{"epoch": 1.21875, "grad_norm": 0.009964634664356709, "learning_rate": 4.2816614569850635e-05, "loss": 7.889936387073249e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00008, "step": 312, "tokens/total": 9439344, "tokens/train_per_sec_per_gpu": 33.21, "tokens/trainable": 142561} +{"epoch": 1.22265625, "grad_norm": 0.22056631743907928, "learning_rate": 4.2537420897897864e-05, "loss": 0.006305779796093702, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00633, "step": 313, "tokens/total": 9469792, "tokens/train_per_sec_per_gpu": 35.85, "tokens/trainable": 143046} +{"epoch": 1.2265625, "grad_norm": 0.024270815774798393, "learning_rate": 4.225874584473174e-05, "loss": 0.0002287090610479936, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00023, "step": 314, "tokens/total": 9500128, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 143493} +{"epoch": 1.23046875, "grad_norm": 0.11968889087438583, "learning_rate": 4.19806010071587e-05, "loss": 0.0012203591177240014, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00122, "step": 315, "tokens/total": 9530480, "tokens/train_per_sec_per_gpu": 35.43, "tokens/trainable": 143956} +{"epoch": 1.234375, "grad_norm": 0.28869929909706116, "learning_rate": 4.170299795992081e-05, "loss": 0.0029480340890586376, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00295, "step": 316, "tokens/total": 9560912, "tokens/train_per_sec_per_gpu": 34.17, "tokens/trainable": 144411} +{"epoch": 1.23828125, "grad_norm": 0.08119015395641327, "learning_rate": 4.142594825521398e-05, "loss": 0.0008730962872505188, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00087, "step": 317, "tokens/total": 9591344, "tokens/train_per_sec_per_gpu": 38.84, "tokens/trainable": 144892} +{"epoch": 1.2421875, "grad_norm": 0.24371854960918427, "learning_rate": 4.114946342220728e-05, "loss": 0.0037464506458491087, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00375, "step": 318, "tokens/total": 9621392, "tokens/train_per_sec_per_gpu": 32.83, "tokens/trainable": 145339} +{"epoch": 1.24609375, "grad_norm": 0.0005952336359769106, "learning_rate": 4.087355496656321e-05, "loss": 1.7879261577036232e-05, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00002, "step": 319, "tokens/total": 9651728, "tokens/train_per_sec_per_gpu": 36.59, "tokens/trainable": 145811} +{"epoch": 1.25, "grad_norm": 0.21333903074264526, "learning_rate": 4.05982343699588e-05, "loss": 0.0022262686397880316, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00223, "step": 320, "tokens/total": 9682224, "tokens/train_per_sec_per_gpu": 38.92, "tokens/trainable": 146314} +{"epoch": 1.25390625, "grad_norm": 0.12917986512184143, "learning_rate": 4.0323513089607876e-05, "loss": 0.0012778663076460361, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00128, "step": 321, "tokens/total": 9712656, "tokens/train_per_sec_per_gpu": 31.93, "tokens/trainable": 146781} +{"epoch": 1.2578125, "grad_norm": 0.011195745319128036, "learning_rate": 4.004940255778431e-05, "loss": 0.00012576431618072093, "memory/device_reserved (GiB)": 34.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00013, "step": 322, "tokens/total": 9743088, "tokens/train_per_sec_per_gpu": 35.69, "tokens/trainable": 147225} +{"epoch": 1.26171875, "grad_norm": 0.11680588871240616, "learning_rate": 3.977591418134619e-05, "loss": 0.001117704901844263, "memory/device_reserved (GiB)": 35.17, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00112, "step": 323, "tokens/total": 9773808, "tokens/train_per_sec_per_gpu": 35.02, "tokens/trainable": 147673} +{"epoch": 1.265625, "grad_norm": 0.1339329034090042, "learning_rate": 3.95030593412612e-05, "loss": 0.0011167675256729126, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00112, "step": 324, "tokens/total": 9804016, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 148103} +{"epoch": 1.26953125, "grad_norm": 0.00400894321501255, "learning_rate": 3.923084939213296e-05, "loss": 5.7869408919941634e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00006, "step": 325, "tokens/total": 9834592, "tokens/train_per_sec_per_gpu": 31.93, "tokens/trainable": 148535} +{"epoch": 1.2734375, "grad_norm": 0.012578712776303291, "learning_rate": 3.895929566172861e-05, "loss": 0.00020866318664047867, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00021, "step": 326, "tokens/total": 9864848, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 148999} +{"epoch": 1.27734375, "grad_norm": 0.4467204213142395, "learning_rate": 3.868840945050728e-05, "loss": 0.013730829581618309, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01383, "step": 327, "tokens/total": 9895168, "tokens/train_per_sec_per_gpu": 31.71, "tokens/trainable": 149431} +{"epoch": 1.28125, "grad_norm": 0.002340462524443865, "learning_rate": 3.841820203114995e-05, "loss": 4.573522164719179e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00005, "step": 328, "tokens/total": 9925696, "tokens/train_per_sec_per_gpu": 34.08, "tokens/trainable": 149886} +{"epoch": 1.28515625, "grad_norm": 0.005373048596084118, "learning_rate": 3.814868464809027e-05, "loss": 5.34585997229442e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00005, "step": 329, "tokens/total": 9955648, "tokens/train_per_sec_per_gpu": 31.06, "tokens/trainable": 150288} +{"epoch": 1.2890625, "grad_norm": 0.8296674489974976, "learning_rate": 3.787986851704667e-05, "loss": 0.009736484847962856, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00978, "step": 330, "tokens/total": 9985856, "tokens/train_per_sec_per_gpu": 34.08, "tokens/trainable": 150733} +{"epoch": 1.29296875, "grad_norm": 0.01640630140900612, "learning_rate": 3.7611764824555654e-05, "loss": 0.00026526901638135314, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00027, "step": 331, "tokens/total": 10016208, "tokens/train_per_sec_per_gpu": 37.03, "tokens/trainable": 151207} +{"epoch": 1.296875, "grad_norm": 0.013102808967232704, "learning_rate": 3.734438472750619e-05, "loss": 0.00013475351443048567, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00013, "step": 332, "tokens/total": 10046512, "tokens/train_per_sec_per_gpu": 37.81, "tokens/trainable": 151694} +{"epoch": 1.30078125, "grad_norm": 0.01519423071295023, "learning_rate": 3.707773935267552e-05, "loss": 0.0002597831189632416, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00026, "step": 333, "tokens/total": 10076944, "tokens/train_per_sec_per_gpu": 38.61, "tokens/trainable": 152188} +{"epoch": 1.3046875, "grad_norm": 0.002785792574286461, "learning_rate": 3.68118397962661e-05, "loss": 5.2896200941177085e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00005, "step": 334, "tokens/total": 10107296, "tokens/train_per_sec_per_gpu": 30.86, "tokens/trainable": 152596} +{"epoch": 1.30859375, "grad_norm": 0.045195769518613815, "learning_rate": 3.654669712344384e-05, "loss": 0.000801014481112361, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0008, "step": 335, "tokens/total": 10137568, "tokens/train_per_sec_per_gpu": 36.79, "tokens/trainable": 153052} +{"epoch": 1.3125, "grad_norm": 0.028165830299258232, "learning_rate": 3.628232236787763e-05, "loss": 0.0004341424209997058, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00043, "step": 336, "tokens/total": 10167952, "tokens/train_per_sec_per_gpu": 30.42, "tokens/trainable": 153491} +{"epoch": 1.31640625, "grad_norm": 0.07202710211277008, "learning_rate": 3.6018726531280144e-05, "loss": 0.0020600268617272377, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00206, "step": 337, "tokens/total": 10198512, "tokens/train_per_sec_per_gpu": 40.33, "tokens/trainable": 153983} +{"epoch": 1.3203125, "grad_norm": 0.05288974568247795, "learning_rate": 3.575592058295017e-05, "loss": 0.0013123506214469671, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00131, "step": 338, "tokens/total": 10228752, "tokens/train_per_sec_per_gpu": 36.7, "tokens/trainable": 154464} +{"epoch": 1.32421875, "grad_norm": 0.0019543056841939688, "learning_rate": 3.549391545931585e-05, "loss": 4.880658889305778e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00005, "step": 339, "tokens/total": 10259104, "tokens/train_per_sec_per_gpu": 36.04, "tokens/trainable": 154924} +{"epoch": 1.328125, "grad_norm": 0.00155684445053339, "learning_rate": 3.5232722063479914e-05, "loss": 4.453617293620482e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00004, "step": 340, "tokens/total": 10289520, "tokens/train_per_sec_per_gpu": 30.93, "tokens/trainable": 155373} +{"epoch": 1.33203125, "grad_norm": 0.012292287312448025, "learning_rate": 3.49723512647657e-05, "loss": 0.00024946065968833864, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00025, "step": 341, "tokens/total": 10320016, "tokens/train_per_sec_per_gpu": 36.39, "tokens/trainable": 155873} +{"epoch": 1.3359375, "grad_norm": 0.008822723291814327, "learning_rate": 3.471281389826491e-05, "loss": 0.00019343552412465215, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00019, "step": 342, "tokens/total": 10350368, "tokens/train_per_sec_per_gpu": 30.53, "tokens/trainable": 156278} +{"epoch": 1.33984375, "grad_norm": 0.8359783291816711, "learning_rate": 3.4454120764386764e-05, "loss": 0.015975475311279297, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0161, "step": 343, "tokens/total": 10380624, "tokens/train_per_sec_per_gpu": 35.59, "tokens/trainable": 156733} +{"epoch": 1.34375, "grad_norm": 0.012664229609072208, "learning_rate": 3.4196282628408526e-05, "loss": 0.00016400852473452687, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00016, "step": 344, "tokens/total": 10411024, "tokens/train_per_sec_per_gpu": 34.09, "tokens/trainable": 157203} +{"epoch": 1.34765625, "grad_norm": 0.01816992089152336, "learning_rate": 3.3939310220027456e-05, "loss": 0.00016030078404583037, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00016, "step": 345, "tokens/total": 10441248, "tokens/train_per_sec_per_gpu": 38.08, "tokens/trainable": 157707} +{"epoch": 1.3515625, "grad_norm": 0.011242903769016266, "learning_rate": 3.3683214232914404e-05, "loss": 0.00022866495419293642, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00023, "step": 346, "tokens/total": 10471712, "tokens/train_per_sec_per_gpu": 35.18, "tokens/trainable": 158180} +{"epoch": 1.35546875, "grad_norm": 0.005438812542706728, "learning_rate": 3.342800532426873e-05, "loss": 0.00012295089254621416, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00012, "step": 347, "tokens/total": 10502288, "tokens/train_per_sec_per_gpu": 34.53, "tokens/trainable": 158646} +{"epoch": 1.359375, "grad_norm": 0.05521470680832863, "learning_rate": 3.317369411437484e-05, "loss": 0.0009212898439727724, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00092, "step": 348, "tokens/total": 10532304, "tokens/train_per_sec_per_gpu": 35.8, "tokens/trainable": 159115} +{"epoch": 1.36328125, "grad_norm": 0.045007552951574326, "learning_rate": 3.292029118616024e-05, "loss": 0.0007900734199211001, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00079, "step": 349, "tokens/total": 10562608, "tokens/train_per_sec_per_gpu": 31.79, "tokens/trainable": 159538} +{"epoch": 1.3671875, "grad_norm": 0.18341247737407684, "learning_rate": 3.266780708475511e-05, "loss": 0.005364480894058943, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00538, "step": 350, "tokens/total": 10592992, "tokens/train_per_sec_per_gpu": 33.29, "tokens/trainable": 160016} +{"epoch": 1.37109375, "grad_norm": 0.0726502537727356, "learning_rate": 3.241625231705354e-05, "loss": 0.001004248857498169, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.001, "step": 351, "tokens/total": 10623328, "tokens/train_per_sec_per_gpu": 34.52, "tokens/trainable": 160475} +{"epoch": 1.375, "grad_norm": 0.013726887293159962, "learning_rate": 3.216563735127618e-05, "loss": 0.00017499152454547584, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00018, "step": 352, "tokens/total": 10653632, "tokens/train_per_sec_per_gpu": 34.16, "tokens/trainable": 160952} +{"epoch": 1.37890625, "grad_norm": 0.15467943251132965, "learning_rate": 3.191597261653475e-05, "loss": 0.0034005024936050177, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00341, "step": 353, "tokens/total": 10684176, "tokens/train_per_sec_per_gpu": 33.13, "tokens/trainable": 161420} +{"epoch": 1.3828125, "grad_norm": 0.07023883610963821, "learning_rate": 3.166726850239794e-05, "loss": 0.0016934729646891356, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00169, "step": 354, "tokens/total": 10714352, "tokens/train_per_sec_per_gpu": 34.33, "tokens/trainable": 161868} +{"epoch": 1.38671875, "grad_norm": 0.002377701224759221, "learning_rate": 3.141953535845912e-05, "loss": 5.6141776440199465e-05, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00006, "step": 355, "tokens/total": 10744464, "tokens/train_per_sec_per_gpu": 35.61, "tokens/trainable": 162318} +{"epoch": 1.390625, "grad_norm": 0.01144740916788578, "learning_rate": 3.11727834939056e-05, "loss": 0.00025334605015814304, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00025, "step": 356, "tokens/total": 10774976, "tokens/train_per_sec_per_gpu": 31.68, "tokens/trainable": 162789} +{"epoch": 1.39453125, "grad_norm": 0.14478880167007446, "learning_rate": 3.092702317708967e-05, "loss": 0.003968201577663422, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00398, "step": 357, "tokens/total": 10805600, "tokens/train_per_sec_per_gpu": 36.54, "tokens/trainable": 163254} +{"epoch": 1.3984375, "grad_norm": 0.022581662982702255, "learning_rate": 3.0682264635101276e-05, "loss": 0.00020885399135295302, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00021, "step": 358, "tokens/total": 10835920, "tokens/train_per_sec_per_gpu": 34.66, "tokens/trainable": 163715} +{"epoch": 1.40234375, "grad_norm": 2.1134932041168213, "learning_rate": 3.0438518053342407e-05, "loss": 0.000105415572761558, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00011, "step": 359, "tokens/total": 10866048, "tokens/train_per_sec_per_gpu": 34.44, "tokens/trainable": 164197} +{"epoch": 1.40625, "grad_norm": 0.0530208945274353, "learning_rate": 3.0195793575103266e-05, "loss": 0.0006053540273569524, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00061, "step": 360, "tokens/total": 10896288, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 164661} +{"epoch": 1.41015625, "grad_norm": 0.032864298671483994, "learning_rate": 2.9954101301140146e-05, "loss": 0.00037546915700659156, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00038, "step": 361, "tokens/total": 10926816, "tokens/train_per_sec_per_gpu": 33.21, "tokens/trainable": 165111} +{"epoch": 1.4140625, "grad_norm": 0.040343064814805984, "learning_rate": 2.9713451289255123e-05, "loss": 0.00035573297645896673, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.31, "memory/max_allocated (GiB)": 33.31, "ppl": 1.00036, "step": 362, "tokens/total": 10954928, "tokens/train_per_sec_per_gpu": 30.54, "tokens/trainable": 165552} +{"epoch": 1.41796875, "grad_norm": 0.009443948976695538, "learning_rate": 2.9473853553877484e-05, "loss": 0.00012895476538687944, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00013, "step": 363, "tokens/total": 10985328, "tokens/train_per_sec_per_gpu": 35.47, "tokens/trainable": 166026} +{"epoch": 1.421875, "grad_norm": 0.20513421297073364, "learning_rate": 2.9235318065647e-05, "loss": 0.0005669078091159463, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00057, "step": 364, "tokens/total": 11015664, "tokens/train_per_sec_per_gpu": 38.14, "tokens/trainable": 166486} +{"epoch": 1.42578125, "grad_norm": 0.005197233986109495, "learning_rate": 2.8997854750998964e-05, "loss": 8.792963490122929e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00009, "step": 365, "tokens/total": 11046256, "tokens/train_per_sec_per_gpu": 36.61, "tokens/trainable": 166959} +{"epoch": 1.4296875, "grad_norm": 0.005207398906350136, "learning_rate": 2.8761473491751258e-05, "loss": 7.358018774539232e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.68, "memory/max_allocated (GiB)": 33.68, "ppl": 1.00007, "step": 366, "tokens/total": 11076288, "tokens/train_per_sec_per_gpu": 33.89, "tokens/trainable": 167394} +{"epoch": 1.43359375, "grad_norm": 0.008117729797959328, "learning_rate": 2.8526184124692883e-05, "loss": 0.00011434487532824278, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00011, "step": 367, "tokens/total": 11106416, "tokens/train_per_sec_per_gpu": 28.52, "tokens/trainable": 167832} +{"epoch": 1.4375, "grad_norm": 0.02638891525566578, "learning_rate": 2.829199644117484e-05, "loss": 0.00033778423676267266, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00034, "step": 368, "tokens/total": 11136848, "tokens/train_per_sec_per_gpu": 30.84, "tokens/trainable": 168242} +{"epoch": 1.44140625, "grad_norm": 0.15596236288547516, "learning_rate": 2.8058920186702553e-05, "loss": 0.00018454823293723166, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00018, "step": 369, "tokens/total": 11167376, "tokens/train_per_sec_per_gpu": 35.83, "tokens/trainable": 168715} +{"epoch": 1.4453125, "grad_norm": 0.013470837846398354, "learning_rate": 2.782696506053033e-05, "loss": 0.0001927075209096074, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00019, "step": 370, "tokens/total": 11197776, "tokens/train_per_sec_per_gpu": 31.96, "tokens/trainable": 169147} +{"epoch": 1.44921875, "grad_norm": 0.009815672412514687, "learning_rate": 2.7596140715257824e-05, "loss": 0.0001360755559289828, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00014, "step": 371, "tokens/total": 11227840, "tokens/train_per_sec_per_gpu": 34.42, "tokens/trainable": 169628} +{"epoch": 1.453125, "grad_norm": 0.003195439698174596, "learning_rate": 2.7366456756428184e-05, "loss": 6.715093331877142e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00007, "step": 372, "tokens/total": 11258176, "tokens/train_per_sec_per_gpu": 35.41, "tokens/trainable": 170085} +{"epoch": 1.45703125, "grad_norm": 0.009018086828291416, "learning_rate": 2.7137922742128486e-05, "loss": 8.873787737684324e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00009, "step": 373, "tokens/total": 11288336, "tokens/train_per_sec_per_gpu": 35.81, "tokens/trainable": 170584} +{"epoch": 1.4609375, "grad_norm": 0.02981177344918251, "learning_rate": 2.691054818259188e-05, "loss": 0.0004891970893368125, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00049, "step": 374, "tokens/total": 11318848, "tokens/train_per_sec_per_gpu": 34.86, "tokens/trainable": 171058} +{"epoch": 1.46484375, "grad_norm": 0.28394025564193726, "learning_rate": 2.6684342539801933e-05, "loss": 0.0027279453352093697, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00273, "step": 375, "tokens/total": 11349168, "tokens/train_per_sec_per_gpu": 34.13, "tokens/trainable": 171518} +{"epoch": 1.46875, "grad_norm": 0.06641732156276703, "learning_rate": 2.645931522709877e-05, "loss": 0.0007971529266797006, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0008, "step": 376, "tokens/total": 11379328, "tokens/train_per_sec_per_gpu": 40.75, "tokens/trainable": 172007} +{"epoch": 1.47265625, "grad_norm": 0.26441285014152527, "learning_rate": 2.6235475608787365e-05, "loss": 0.0031230500899255276, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00313, "step": 377, "tokens/total": 11409568, "tokens/train_per_sec_per_gpu": 34.91, "tokens/trainable": 172468} +{"epoch": 1.4765625, "grad_norm": 0.0731891617178917, "learning_rate": 2.6012832999747916e-05, "loss": 0.0010390954557806253, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00104, "step": 378, "tokens/total": 11439968, "tokens/train_per_sec_per_gpu": 33.63, "tokens/trainable": 172923} +{"epoch": 1.48046875, "grad_norm": 0.32646894454956055, "learning_rate": 2.579139666504821e-05, "loss": 0.011534517630934715, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0116, "step": 379, "tokens/total": 11470496, "tokens/train_per_sec_per_gpu": 34.28, "tokens/trainable": 173370} +{"epoch": 1.484375, "grad_norm": 0.020526761189103127, "learning_rate": 2.557117581955798e-05, "loss": 0.0003196616598870605, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00032, "step": 380, "tokens/total": 11500720, "tokens/train_per_sec_per_gpu": 34.58, "tokens/trainable": 173837} +{"epoch": 1.48828125, "grad_norm": 0.07209137082099915, "learning_rate": 2.5352179627565532e-05, "loss": 0.0009562239283695817, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00096, "step": 381, "tokens/total": 11531280, "tokens/train_per_sec_per_gpu": 35.79, "tokens/trainable": 174294} +{"epoch": 1.4921875, "grad_norm": 0.015863383188843727, "learning_rate": 2.5134417202396277e-05, "loss": 0.00018306101264897734, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00018, "step": 382, "tokens/total": 11561264, "tokens/train_per_sec_per_gpu": 30.99, "tokens/trainable": 174701} +{"epoch": 1.49609375, "grad_norm": 0.031188542023301125, "learning_rate": 2.491789760603361e-05, "loss": 0.0005925485165789723, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00059, "step": 383, "tokens/total": 11591616, "tokens/train_per_sec_per_gpu": 32.26, "tokens/trainable": 175131} +{"epoch": 1.5, "grad_norm": 0.03409359231591225, "learning_rate": 2.4702629848741764e-05, "loss": 0.0006768744206055999, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00068, "step": 384, "tokens/total": 11622000, "tokens/train_per_sec_per_gpu": 34.18, "tokens/trainable": 175586} +{"epoch": 1.50390625, "grad_norm": 0.022114347666502, "learning_rate": 2.4488622888690785e-05, "loss": 0.00034674766357056797, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00035, "step": 385, "tokens/total": 11652352, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 176026} +{"epoch": 1.5078125, "grad_norm": 0.007172535173594952, "learning_rate": 2.427588563158384e-05, "loss": 0.00010547737474553287, "memory/device_reserved (GiB)": 34.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00011, "step": 386, "tokens/total": 11682736, "tokens/train_per_sec_per_gpu": 33.91, "tokens/trainable": 176512} +{"epoch": 1.51171875, "grad_norm": 0.011641777120530605, "learning_rate": 2.406442693028651e-05, "loss": 0.00017262669280171394, "memory/device_reserved (GiB)": 34.98, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00017, "step": 387, "tokens/total": 11713136, "tokens/train_per_sec_per_gpu": 30.39, "tokens/trainable": 176943} +{"epoch": 1.515625, "grad_norm": 0.0061674294993281364, "learning_rate": 2.3854255584458547e-05, "loss": 9.34262279770337e-05, "memory/device_reserved (GiB)": 35.23, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00009, "step": 388, "tokens/total": 11743808, "tokens/train_per_sec_per_gpu": 32.21, "tokens/trainable": 177370} +{"epoch": 1.51953125, "grad_norm": 0.04236450046300888, "learning_rate": 2.3645380340187508e-05, "loss": 0.0003042269090656191, "memory/device_reserved (GiB)": 35.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0003, "step": 389, "tokens/total": 11774144, "tokens/train_per_sec_per_gpu": 35.32, "tokens/trainable": 177851} +{"epoch": 1.5234375, "grad_norm": 0.013182509690523148, "learning_rate": 2.3437809889624914e-05, "loss": 0.00026280959718860686, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00026, "step": 390, "tokens/total": 11804272, "tokens/train_per_sec_per_gpu": 34.9, "tokens/trainable": 178310} +{"epoch": 1.52734375, "grad_norm": 0.012172302231192589, "learning_rate": 2.3231552870624487e-05, "loss": 8.891556353773922e-05, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00009, "step": 391, "tokens/total": 11832432, "tokens/train_per_sec_per_gpu": 37.01, "tokens/trainable": 178736} +{"epoch": 1.53125, "grad_norm": 0.09926214069128036, "learning_rate": 2.3026617866382657e-05, "loss": 0.0020971111953258514, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0021, "step": 392, "tokens/total": 11862960, "tokens/train_per_sec_per_gpu": 36.43, "tokens/trainable": 179233} +{"epoch": 1.53515625, "grad_norm": 0.006282840855419636, "learning_rate": 2.2823013405081507e-05, "loss": 0.00012446771143004298, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00012, "step": 393, "tokens/total": 11893232, "tokens/train_per_sec_per_gpu": 38.72, "tokens/trainable": 179730} +{"epoch": 1.5390625, "grad_norm": 0.012727927416563034, "learning_rate": 2.2620747959533722e-05, "loss": 0.0001671911304583773, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00017, "step": 394, "tokens/total": 11923664, "tokens/train_per_sec_per_gpu": 38.03, "tokens/trainable": 180227} +{"epoch": 1.54296875, "grad_norm": 0.007153753656893969, "learning_rate": 2.2419829946830123e-05, "loss": 0.00010278529953211546, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0001, "step": 395, "tokens/total": 11954032, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 180687} +{"epoch": 1.546875, "grad_norm": 0.026691973209381104, "learning_rate": 2.2220267727989325e-05, "loss": 0.00036194472340866923, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00036, "step": 396, "tokens/total": 11984512, "tokens/train_per_sec_per_gpu": 32.85, "tokens/trainable": 181141} +{"epoch": 1.55078125, "grad_norm": 0.1592523753643036, "learning_rate": 2.202206960760984e-05, "loss": 0.0023322016932070255, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00233, "step": 397, "tokens/total": 12014880, "tokens/train_per_sec_per_gpu": 35.94, "tokens/trainable": 181648} +{"epoch": 1.5546875, "grad_norm": 0.02613229863345623, "learning_rate": 2.182524383352446e-05, "loss": 0.0003333169443067163, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00033, "step": 398, "tokens/total": 12044928, "tokens/train_per_sec_per_gpu": 33.92, "tokens/trainable": 182109} +{"epoch": 1.55859375, "grad_norm": 0.006405193824321032, "learning_rate": 2.1629798596457056e-05, "loss": 8.373583841603249e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00008, "step": 399, "tokens/total": 12075344, "tokens/train_per_sec_per_gpu": 31.7, "tokens/trainable": 182559} +{"epoch": 1.5625, "grad_norm": 0.01651483029127121, "learning_rate": 2.1435742029681725e-05, "loss": 0.0001667532924329862, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00017, "step": 400, "tokens/total": 12105616, "tokens/train_per_sec_per_gpu": 33.32, "tokens/trainable": 182985} +{"epoch": 1.56640625, "grad_norm": 0.0015016240067780018, "learning_rate": 2.124308220868431e-05, "loss": 3.507245855871588e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00004, "step": 401, "tokens/total": 12136080, "tokens/train_per_sec_per_gpu": 32.89, "tokens/trainable": 183440} +{"epoch": 1.5703125, "grad_norm": 0.03716011717915535, "learning_rate": 2.105182715082638e-05, "loss": 0.0006280053639784455, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00063, "step": 402, "tokens/total": 12166240, "tokens/train_per_sec_per_gpu": 34.66, "tokens/trainable": 183917} +{"epoch": 1.57421875, "grad_norm": 0.0026928475126624107, "learning_rate": 2.0861984815011552e-05, "loss": 6.0345602832967415e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00006, "step": 403, "tokens/total": 12196544, "tokens/train_per_sec_per_gpu": 30.51, "tokens/trainable": 184348} +{"epoch": 1.578125, "grad_norm": 0.011999940499663353, "learning_rate": 2.0673563101354323e-05, "loss": 0.00010794817353598773, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00011, "step": 404, "tokens/total": 12226784, "tokens/train_per_sec_per_gpu": 31.7, "tokens/trainable": 184786} +{"epoch": 1.58203125, "grad_norm": 0.003504133317619562, "learning_rate": 2.0486569850851317e-05, "loss": 4.769703446072526e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00005, "step": 405, "tokens/total": 12257264, "tokens/train_per_sec_per_gpu": 33.99, "tokens/trainable": 185249} +{"epoch": 1.5859375, "grad_norm": 0.0031612543389201164, "learning_rate": 2.0301012845054956e-05, "loss": 5.125769166625105e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00005, "step": 406, "tokens/total": 12287584, "tokens/train_per_sec_per_gpu": 31.66, "tokens/trainable": 185680} +{"epoch": 1.58984375, "grad_norm": 0.005299390759319067, "learning_rate": 2.011689980574966e-05, "loss": 7.910770364105701e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00008, "step": 407, "tokens/total": 12317984, "tokens/train_per_sec_per_gpu": 31.88, "tokens/trainable": 186100} +{"epoch": 1.59375, "grad_norm": 0.003536214353516698, "learning_rate": 1.993423839463052e-05, "loss": 6.525318895000964e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00007, "step": 408, "tokens/total": 12348144, "tokens/train_per_sec_per_gpu": 40.05, "tokens/trainable": 186565} +{"epoch": 1.59765625, "grad_norm": 0.011145357973873615, "learning_rate": 1.975303621298445e-05, "loss": 0.00013281722203828394, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00013, "step": 409, "tokens/total": 12378544, "tokens/train_per_sec_per_gpu": 36.04, "tokens/trainable": 187017} +{"epoch": 1.6015625, "grad_norm": 0.004464628174901009, "learning_rate": 1.957330080137385e-05, "loss": 8.551983773941174e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00009, "step": 410, "tokens/total": 12408784, "tokens/train_per_sec_per_gpu": 33.21, "tokens/trainable": 187468} +{"epoch": 1.60546875, "grad_norm": 0.004235507920384407, "learning_rate": 1.9395039639322864e-05, "loss": 7.31897889636457e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00007, "step": 411, "tokens/total": 12438832, "tokens/train_per_sec_per_gpu": 37.81, "tokens/trainable": 187961} +{"epoch": 1.609375, "grad_norm": 0.0010779709555208683, "learning_rate": 1.9218260145006073e-05, "loss": 1.925879405462183e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00002, "step": 412, "tokens/total": 12469296, "tokens/train_per_sec_per_gpu": 36.12, "tokens/trainable": 188447} +{"epoch": 1.61328125, "grad_norm": 0.043327320367097855, "learning_rate": 1.904296967493982e-05, "loss": 0.0004905068781226873, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00049, "step": 413, "tokens/total": 12499392, "tokens/train_per_sec_per_gpu": 36.17, "tokens/trainable": 188903} +{"epoch": 1.6171875, "grad_norm": 0.13874047994613647, "learning_rate": 1.8869175523676064e-05, "loss": 0.0003094755229540169, "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": 36.7, "tokens/trainable": 189398} +{"epoch": 1.62109375, "grad_norm": 0.0014462472172454, "learning_rate": 1.869688492349885e-05, "loss": 3.0223447538446635e-05, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00003, "step": 415, "tokens/total": 12560240, "tokens/train_per_sec_per_gpu": 34.33, "tokens/trainable": 189850} +{"epoch": 1.625, "grad_norm": 0.0014046201249584556, "learning_rate": 1.85261050441233e-05, "loss": 2.9643942980328575e-05, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00003, "step": 416, "tokens/total": 12590736, "tokens/train_per_sec_per_gpu": 32.86, "tokens/trainable": 190306} +{"epoch": 1.62890625, "grad_norm": 0.000552138953935355, "learning_rate": 1.8356842992397304e-05, "loss": 1.5713212633272633e-05, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00002, "step": 417, "tokens/total": 12621168, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 190746} +{"epoch": 1.6328125, "grad_norm": 0.005376334302127361, "learning_rate": 1.8189105812005714e-05, "loss": 6.261414819164202e-05, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00006, "step": 418, "tokens/total": 12651456, "tokens/train_per_sec_per_gpu": 31.31, "tokens/trainable": 191156} +{"epoch": 1.63671875, "grad_norm": 0.00120884261559695, "learning_rate": 1.802290048317732e-05, "loss": 2.686155858100392e-05, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00003, "step": 419, "tokens/total": 12681760, "tokens/train_per_sec_per_gpu": 29.64, "tokens/trainable": 191573} +{"epoch": 1.640625, "grad_norm": 0.004546779207885265, "learning_rate": 1.785823392239424e-05, "loss": 7.581210957141593e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00008, "step": 420, "tokens/total": 12712144, "tokens/train_per_sec_per_gpu": 32.92, "tokens/trainable": 192014} +{"epoch": 1.64453125, "grad_norm": 0.014719134196639061, "learning_rate": 1.7695112982104225e-05, "loss": 0.00021209896658547223, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00021, "step": 421, "tokens/total": 12742400, "tokens/train_per_sec_per_gpu": 33.02, "tokens/trainable": 192453} +{"epoch": 1.6484375, "grad_norm": 0.0006049374933354557, "learning_rate": 1.7533544450435433e-05, "loss": 1.7317186575382948e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00002, "step": 422, "tokens/total": 12772288, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 192886} +{"epoch": 1.65234375, "grad_norm": 0.01476835086941719, "learning_rate": 1.7373535050913946e-05, "loss": 0.0001859527692431584, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00019, "step": 423, "tokens/total": 12802880, "tokens/train_per_sec_per_gpu": 30.05, "tokens/trainable": 193323} +{"epoch": 1.65625, "grad_norm": 0.003950290847569704, "learning_rate": 1.721509144218405e-05, "loss": 9.087299986276776e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00009, "step": 424, "tokens/total": 12833456, "tokens/train_per_sec_per_gpu": 32.83, "tokens/trainable": 193753} +{"epoch": 1.66015625, "grad_norm": 0.008329745382070541, "learning_rate": 1.705822021773101e-05, "loss": 8.313520083902404e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00008, "step": 425, "tokens/total": 12863952, "tokens/train_per_sec_per_gpu": 36.61, "tokens/trainable": 194240} +{"epoch": 1.6640625, "grad_norm": 0.5991763472557068, "learning_rate": 1.69029279056068e-05, "loss": 0.004271846264600754, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00428, "step": 426, "tokens/total": 12894352, "tokens/train_per_sec_per_gpu": 35.63, "tokens/trainable": 194680} +{"epoch": 1.66796875, "grad_norm": 0.02893899939954281, "learning_rate": 1.6749220968158415e-05, "loss": 0.00015647552208974957, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00016, "step": 427, "tokens/total": 12924688, "tokens/train_per_sec_per_gpu": 38.65, "tokens/trainable": 195186} +{"epoch": 1.671875, "grad_norm": 0.0829748809337616, "learning_rate": 1.659710580175893e-05, "loss": 0.0008744620135985315, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00087, "step": 428, "tokens/total": 12955072, "tokens/train_per_sec_per_gpu": 36.69, "tokens/trainable": 195693} +{"epoch": 1.67578125, "grad_norm": 0.00444152532145381, "learning_rate": 1.644658873654133e-05, "loss": 6.830539496149868e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00007, "step": 429, "tokens/total": 12985328, "tokens/train_per_sec_per_gpu": 37.21, "tokens/trainable": 196130} +{"epoch": 1.6796875, "grad_norm": 0.02823544666171074, "learning_rate": 1.629767603613508e-05, "loss": 0.0002984766324516386, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.0003, "step": 430, "tokens/total": 13015600, "tokens/train_per_sec_per_gpu": 29.76, "tokens/trainable": 196530} +{"epoch": 1.68359375, "grad_norm": 6.060963153839111, "learning_rate": 1.615037389740547e-05, "loss": 0.00036459555849432945, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00036, "step": 431, "tokens/total": 13045728, "tokens/train_per_sec_per_gpu": 30.12, "tokens/trainable": 196974} +{"epoch": 1.6875, "grad_norm": 0.01055138185620308, "learning_rate": 1.600468845019576e-05, "loss": 0.00012494996190071106, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00012, "step": 432, "tokens/total": 13075920, "tokens/train_per_sec_per_gpu": 35.5, "tokens/trainable": 197454} +{"epoch": 1.69140625, "grad_norm": 0.00017622209270484746, "learning_rate": 1.5860625757072092e-05, "loss": 6.37679568171734e-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": 31.6, "tokens/trainable": 197902} +{"epoch": 1.6953125, "grad_norm": 0.04205000028014183, "learning_rate": 1.571819181307116e-05, "loss": 0.00029261858435347676, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00029, "step": 434, "tokens/total": 13136368, "tokens/train_per_sec_per_gpu": 36.27, "tokens/trainable": 198374} +{"epoch": 1.69921875, "grad_norm": 0.012470721267163754, "learning_rate": 1.557739254545075e-05, "loss": 0.00017825230315793306, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00018, "step": 435, "tokens/total": 13166704, "tokens/train_per_sec_per_gpu": 37.77, "tokens/trainable": 198846} +{"epoch": 1.703125, "grad_norm": 0.002119631739333272, "learning_rate": 1.543823381344311e-05, "loss": 3.121166810160503e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00003, "step": 436, "tokens/total": 13197344, "tokens/train_per_sec_per_gpu": 37.82, "tokens/trainable": 199320} +{"epoch": 1.70703125, "grad_norm": 0.005058391951024532, "learning_rate": 1.5300721408011114e-05, "loss": 5.037401570007205e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00005, "step": 437, "tokens/total": 13227664, "tokens/train_per_sec_per_gpu": 37.58, "tokens/trainable": 199777} +{"epoch": 1.7109375, "grad_norm": 0.0023086639121174812, "learning_rate": 1.5164861051607254e-05, "loss": 3.312892295070924e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 438, "tokens/total": 13257632, "tokens/train_per_sec_per_gpu": 31.13, "tokens/trainable": 200194} +{"epoch": 1.71484375, "grad_norm": 0.0005081371637061238, "learning_rate": 1.5030658397935521e-05, "loss": 1.3808794392389245e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00001, "step": 439, "tokens/total": 13288240, "tokens/train_per_sec_per_gpu": 34.06, "tokens/trainable": 200663} +{"epoch": 1.71875, "grad_norm": 0.03984901309013367, "learning_rate": 1.4898119031716104e-05, "loss": 0.00027867406606674194, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00028, "step": 440, "tokens/total": 13318992, "tokens/train_per_sec_per_gpu": 34.05, "tokens/trainable": 201121} +{"epoch": 1.72265625, "grad_norm": 0.004065736196935177, "learning_rate": 1.476724846845306e-05, "loss": 4.005522350780666e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00004, "step": 441, "tokens/total": 13349392, "tokens/train_per_sec_per_gpu": 40.07, "tokens/trainable": 201598} +{"epoch": 1.7265625, "grad_norm": 0.0009828282054513693, "learning_rate": 1.463805215420471e-05, "loss": 2.004339148697909e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 442, "tokens/total": 13379648, "tokens/train_per_sec_per_gpu": 35.29, "tokens/trainable": 202068} +{"epoch": 1.73046875, "grad_norm": 0.0016327777411788702, "learning_rate": 1.451053546535705e-05, "loss": 2.840825436578598e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00003, "step": 443, "tokens/total": 13409936, "tokens/train_per_sec_per_gpu": 35.78, "tokens/trainable": 202524} +{"epoch": 1.734375, "grad_norm": 0.07271858304738998, "learning_rate": 1.438470370840001e-05, "loss": 0.00016209090244956315, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00016, "step": 444, "tokens/total": 13440288, "tokens/train_per_sec_per_gpu": 36.54, "tokens/trainable": 202996} +{"epoch": 1.73828125, "grad_norm": 0.0013328593922778964, "learning_rate": 1.4260562119706606e-05, "loss": 2.0676230633398518e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00002, "step": 445, "tokens/total": 13468880, "tokens/train_per_sec_per_gpu": 39.34, "tokens/trainable": 203470} +{"epoch": 1.7421875, "grad_norm": 0.0807131826877594, "learning_rate": 1.413811586531508e-05, "loss": 0.0006098271114751697, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00061, "step": 446, "tokens/total": 13499184, "tokens/train_per_sec_per_gpu": 35.55, "tokens/trainable": 203919} +{"epoch": 1.74609375, "grad_norm": 0.010202539153397083, "learning_rate": 1.4017370040713884e-05, "loss": 0.00011388547136448324, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00011, "step": 447, "tokens/total": 13529632, "tokens/train_per_sec_per_gpu": 33.54, "tokens/trainable": 204376} +{"epoch": 1.75, "grad_norm": 0.012697691097855568, "learning_rate": 1.3898329670629645e-05, "loss": 6.711178866680712e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00007, "step": 448, "tokens/total": 13560080, "tokens/train_per_sec_per_gpu": 32.23, "tokens/trainable": 204821} +{"epoch": 1.75390625, "grad_norm": 0.00029595353407785296, "learning_rate": 1.3780999708818058e-05, "loss": 8.797971531748772e-06, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00001, "step": 449, "tokens/total": 13590496, "tokens/train_per_sec_per_gpu": 33.82, "tokens/trainable": 205271} +{"epoch": 1.7578125, "grad_norm": 0.022445866838097572, "learning_rate": 1.3665385037857758e-05, "loss": 0.00015267175331246108, "memory/device_reserved (GiB)": 34.92, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00015, "step": 450, "tokens/total": 13620960, "tokens/train_per_sec_per_gpu": 31.15, "tokens/trainable": 205723} +{"epoch": 1.76171875, "grad_norm": 0.0014366039540618658, "learning_rate": 1.3551490468947126e-05, "loss": 2.87473012576811e-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": 32.65, "tokens/trainable": 206169} +{"epoch": 1.765625, "grad_norm": 0.0042471266351640224, "learning_rate": 1.3439320741704075e-05, "loss": 5.8180645282845944e-05, "memory/device_reserved (GiB)": 35.04, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00006, "step": 452, "tokens/total": 13681792, "tokens/train_per_sec_per_gpu": 32.96, "tokens/trainable": 206592} +{"epoch": 1.76953125, "grad_norm": 0.009985425509512424, "learning_rate": 1.3328880523968808e-05, "loss": 0.0001178122911369428, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00012, "step": 453, "tokens/total": 13712256, "tokens/train_per_sec_per_gpu": 33.21, "tokens/trainable": 207055} +{"epoch": 1.7734375, "grad_norm": 0.024878855794668198, "learning_rate": 1.3220174411609587e-05, "loss": 0.00022001308389008045, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00022, "step": 454, "tokens/total": 13742128, "tokens/train_per_sec_per_gpu": 34.68, "tokens/trainable": 207494} +{"epoch": 1.77734375, "grad_norm": 0.002308261813595891, "learning_rate": 1.3113206928331471e-05, "loss": 2.7343181500327773e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00003, "step": 455, "tokens/total": 13772160, "tokens/train_per_sec_per_gpu": 33.16, "tokens/trainable": 207956} +{"epoch": 1.78125, "grad_norm": 0.0006800597766414285, "learning_rate": 1.300798252548806e-05, "loss": 1.6906453311094083e-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": 35.72, "tokens/trainable": 208408} +{"epoch": 1.78515625, "grad_norm": 0.04438555985689163, "learning_rate": 1.2904505581896265e-05, "loss": 0.0004473269509617239, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00045, "step": 457, "tokens/total": 13832880, "tokens/train_per_sec_per_gpu": 36.27, "tokens/trainable": 208889} +{"epoch": 1.7890625, "grad_norm": 0.04841502383351326, "learning_rate": 1.2802780403654082e-05, "loss": 0.00028527071117423475, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00029, "step": 458, "tokens/total": 13862880, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 209336} +{"epoch": 1.79296875, "grad_norm": 0.011998557485640049, "learning_rate": 1.2702811223961408e-05, "loss": 0.00014353993174154311, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00014, "step": 459, "tokens/total": 13893440, "tokens/train_per_sec_per_gpu": 33.51, "tokens/trainable": 209797} +{"epoch": 1.796875, "grad_norm": 0.0004317575367167592, "learning_rate": 1.2604602202943861e-05, "loss": 1.386312760587316e-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": 36.34, "tokens/trainable": 210240} +{"epoch": 1.80078125, "grad_norm": 0.13330896198749542, "learning_rate": 1.2508157427479686e-05, "loss": 0.001426919479854405, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00143, "step": 461, "tokens/total": 13954208, "tokens/train_per_sec_per_gpu": 35.15, "tokens/trainable": 210702} +{"epoch": 1.8046875, "grad_norm": 0.015856783837080002, "learning_rate": 1.2413480911029655e-05, "loss": 0.00015326723223552108, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00015, "step": 462, "tokens/total": 13984784, "tokens/train_per_sec_per_gpu": 31.75, "tokens/trainable": 211140} +{"epoch": 1.80859375, "grad_norm": 0.0002883273409679532, "learning_rate": 1.2320576593470082e-05, "loss": 8.843874638841953e-06, "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": 32.65, "tokens/trainable": 211575} +{"epoch": 1.8125, "grad_norm": 0.012603677809238434, "learning_rate": 1.2229448340928828e-05, "loss": 6.793404463678598e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00007, "step": 464, "tokens/total": 14045312, "tokens/train_per_sec_per_gpu": 39.02, "tokens/trainable": 212083} +{"epoch": 1.81640625, "grad_norm": 0.00318150850944221, "learning_rate": 1.2140099945624458e-05, "loss": 4.058713966514915e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00004, "step": 465, "tokens/total": 14075840, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 212558} +{"epoch": 1.8203125, "grad_norm": 0.0012220257194712758, "learning_rate": 1.205253512570841e-05, "loss": 2.0981166017008945e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 466, "tokens/total": 14106016, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 213014} +{"epoch": 1.82421875, "grad_norm": 0.0007184451096691191, "learning_rate": 1.1966757525110255e-05, "loss": 1.6628566299914382e-05, "memory/device_reserved (GiB)": 37.99, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00002, "step": 467, "tokens/total": 14136512, "tokens/train_per_sec_per_gpu": 34.24, "tokens/trainable": 213465} +{"epoch": 1.828125, "grad_norm": 0.021840877830982208, "learning_rate": 1.1882770713386095e-05, "loss": 0.00020292821864131838, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0002, "step": 468, "tokens/total": 14166832, "tokens/train_per_sec_per_gpu": 36.87, "tokens/trainable": 213940} +{"epoch": 1.83203125, "grad_norm": 0.017776230350136757, "learning_rate": 1.180057818556998e-05, "loss": 0.00021284582908265293, "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": 35.28, "tokens/trainable": 214415} +{"epoch": 1.8359375, "grad_norm": 0.0011880530510097742, "learning_rate": 1.1720183362028494e-05, "loss": 1.4444680346059613e-05, "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": 34.48, "tokens/trainable": 214880} +{"epoch": 1.83984375, "grad_norm": 0.0028076171875, "learning_rate": 1.1641589588318387e-05, "loss": 2.9122196792741306e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00003, "step": 471, "tokens/total": 14257888, "tokens/train_per_sec_per_gpu": 33.67, "tokens/trainable": 215334} +{"epoch": 1.84375, "grad_norm": 0.00019811835954897106, "learning_rate": 1.1564800135047418e-05, "loss": 6.9107745730434544e-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": 35.0, "tokens/trainable": 215822} +{"epoch": 1.84765625, "grad_norm": 0.0014798284973949194, "learning_rate": 1.148981819773816e-05, "loss": 1.5123480807233136e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 473, "tokens/total": 14318448, "tokens/train_per_sec_per_gpu": 30.95, "tokens/trainable": 216239} +{"epoch": 1.8515625, "grad_norm": 0.026936255395412445, "learning_rate": 1.1416646896695086e-05, "loss": 0.0001889157429104671, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00019, "step": 474, "tokens/total": 14348912, "tokens/train_per_sec_per_gpu": 33.47, "tokens/trainable": 216691} +{"epoch": 1.85546875, "grad_norm": 0.03738206997513771, "learning_rate": 1.1345289276874717e-05, "loss": 0.00048619165318086743, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00049, "step": 475, "tokens/total": 14379440, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 217162} +{"epoch": 1.859375, "grad_norm": 0.03628401830792427, "learning_rate": 1.1275748307758873e-05, "loss": 0.0003041466698050499, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.0003, "step": 476, "tokens/total": 14409840, "tokens/train_per_sec_per_gpu": 36.11, "tokens/trainable": 217638} +{"epoch": 1.86328125, "grad_norm": 0.0003053894324693829, "learning_rate": 1.1208026883231147e-05, "loss": 8.800923751550727e-06, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00001, "step": 477, "tokens/total": 14440256, "tokens/train_per_sec_per_gpu": 33.0, "tokens/trainable": 218053} +{"epoch": 1.8671875, "grad_norm": 0.00386667437851429, "learning_rate": 1.1142127821456433e-05, "loss": 5.062613490736112e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00005, "step": 478, "tokens/total": 14470704, "tokens/train_per_sec_per_gpu": 34.37, "tokens/trainable": 218503} +{"epoch": 1.87109375, "grad_norm": 0.0006709938170388341, "learning_rate": 1.1078053864763674e-05, "loss": 1.1225674825254828e-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": 36.41, "tokens/trainable": 218964} +{"epoch": 1.875, "grad_norm": 0.009718201123178005, "learning_rate": 1.1015807679531756e-05, "loss": 8.914987847674638e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00009, "step": 480, "tokens/total": 14531200, "tokens/train_per_sec_per_gpu": 35.75, "tokens/trainable": 219412} +{"epoch": 1.87890625, "grad_norm": 0.006988356821238995, "learning_rate": 1.0955391856078528e-05, "loss": 0.00010668298637028784, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00011, "step": 481, "tokens/total": 14561152, "tokens/train_per_sec_per_gpu": 38.32, "tokens/trainable": 219886} +{"epoch": 1.8828125, "grad_norm": 0.3569449782371521, "learning_rate": 1.0896808908553007e-05, "loss": 0.002299799118191004, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.0023, "step": 482, "tokens/total": 14591264, "tokens/train_per_sec_per_gpu": 34.75, "tokens/trainable": 220343} +{"epoch": 1.88671875, "grad_norm": 0.09797399491071701, "learning_rate": 1.0840061274830763e-05, "loss": 0.0006474042893387377, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00065, "step": 483, "tokens/total": 14621568, "tokens/train_per_sec_per_gpu": 35.48, "tokens/trainable": 220825} +{"epoch": 1.890625, "grad_norm": 0.018185529857873917, "learning_rate": 1.0785151316412473e-05, "loss": 0.00016680543194524944, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00017, "step": 484, "tokens/total": 14652064, "tokens/train_per_sec_per_gpu": 32.72, "tokens/trainable": 221282} +{"epoch": 1.89453125, "grad_norm": 0.0035336960572749376, "learning_rate": 1.0732081318325639e-05, "loss": 2.4332513930858113e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00002, "step": 485, "tokens/total": 14682608, "tokens/train_per_sec_per_gpu": 38.13, "tokens/trainable": 221780} +{"epoch": 1.8984375, "grad_norm": 0.010691395960748196, "learning_rate": 1.0680853489029501e-05, "loss": 6.945780478417873e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00007, "step": 486, "tokens/total": 14713168, "tokens/train_per_sec_per_gpu": 31.71, "tokens/trainable": 222210} +{"epoch": 1.90234375, "grad_norm": 0.18848194181919098, "learning_rate": 1.0631469960323152e-05, "loss": 0.002461689990013838, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00246, "step": 487, "tokens/total": 14743536, "tokens/train_per_sec_per_gpu": 29.55, "tokens/trainable": 222646} +{"epoch": 1.90625, "grad_norm": 0.007101303432136774, "learning_rate": 1.0583932787256783e-05, "loss": 5.785077519249171e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00006, "step": 488, "tokens/total": 14773728, "tokens/train_per_sec_per_gpu": 33.26, "tokens/trainable": 223066} +{"epoch": 1.91015625, "grad_norm": 0.0015545859932899475, "learning_rate": 1.0538243948046206e-05, "loss": 2.3247019271366298e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00002, "step": 489, "tokens/total": 14803744, "tokens/train_per_sec_per_gpu": 34.39, "tokens/trainable": 223512} +{"epoch": 1.9140625, "grad_norm": 0.00023431811132468283, "learning_rate": 1.0494405343990523e-05, "loss": 6.943540029169526e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00001, "step": 490, "tokens/total": 14832112, "tokens/train_per_sec_per_gpu": 39.18, "tokens/trainable": 223956} +{"epoch": 1.91796875, "grad_norm": 0.015353985130786896, "learning_rate": 1.0452418799392985e-05, "loss": 0.00015685766993556172, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.44, "memory/max_allocated (GiB)": 33.44, "ppl": 1.00016, "step": 491, "tokens/total": 14860560, "tokens/train_per_sec_per_gpu": 36.46, "tokens/trainable": 224416} +{"epoch": 1.921875, "grad_norm": 0.24176718294620514, "learning_rate": 1.0412286061485102e-05, "loss": 0.0033036654349416494, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00331, "step": 492, "tokens/total": 14890848, "tokens/train_per_sec_per_gpu": 34.25, "tokens/trainable": 224883} +{"epoch": 1.92578125, "grad_norm": 0.0026042386889457703, "learning_rate": 1.03740088003539e-05, "loss": 2.109831984853372e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00002, "step": 493, "tokens/total": 14921008, "tokens/train_per_sec_per_gpu": 35.62, "tokens/trainable": 225324} +{"epoch": 1.9296875, "grad_norm": 0.007053546607494354, "learning_rate": 1.0337588608872463e-05, "loss": 5.522219726117328e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00006, "step": 494, "tokens/total": 14951360, "tokens/train_per_sec_per_gpu": 35.12, "tokens/trainable": 225783} +{"epoch": 1.93359375, "grad_norm": 0.0011394162429496646, "learning_rate": 1.0303027002633622e-05, "loss": 1.5214816812658682e-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": 36.6, "tokens/trainable": 226281} +{"epoch": 1.9375, "grad_norm": 0.00018126008217222989, "learning_rate": 1.0270325419886884e-05, "loss": 5.4452602853416465e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00001, "step": 496, "tokens/total": 15011584, "tokens/train_per_sec_per_gpu": 36.96, "tokens/trainable": 226735} +{"epoch": 1.94140625, "grad_norm": 0.0014420925872400403, "learning_rate": 1.0239485221478599e-05, "loss": 2.52262379945023e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00003, "step": 497, "tokens/total": 15042016, "tokens/train_per_sec_per_gpu": 36.55, "tokens/trainable": 227198} +{"epoch": 1.9453125, "grad_norm": 0.0018572796834632754, "learning_rate": 1.0210507690795292e-05, "loss": 3.095180727541447e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.68, "memory/max_allocated (GiB)": 33.68, "ppl": 1.00003, "step": 498, "tokens/total": 15071984, "tokens/train_per_sec_per_gpu": 31.24, "tokens/trainable": 227623} +{"epoch": 1.94921875, "grad_norm": 0.13491077721118927, "learning_rate": 1.0183394033710305e-05, "loss": 0.0014811853179708123, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00148, "step": 499, "tokens/total": 15102192, "tokens/train_per_sec_per_gpu": 32.34, "tokens/trainable": 228074} +{"epoch": 1.953125, "grad_norm": 0.0009476295090280473, "learning_rate": 1.0158145378533583e-05, "loss": 1.7533122445456684e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00002, "step": 500, "tokens/total": 15132560, "tokens/train_per_sec_per_gpu": 35.85, "tokens/trainable": 228541} +{"epoch": 1.95703125, "grad_norm": 0.00058976118452847, "learning_rate": 1.0134762775964726e-05, "loss": 1.0573488907539286e-05, "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": 38.02, "tokens/trainable": 229031} +{"epoch": 1.9609375, "grad_norm": 0.0004935197648592293, "learning_rate": 1.0113247199049278e-05, "loss": 1.0224689503957052e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00001, "step": 502, "tokens/total": 15193456, "tokens/train_per_sec_per_gpu": 38.92, "tokens/trainable": 229512} +{"epoch": 1.96484375, "grad_norm": 0.020853139460086823, "learning_rate": 1.0093599543138205e-05, "loss": 0.00014701965847052634, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00015, "step": 503, "tokens/total": 15223680, "tokens/train_per_sec_per_gpu": 34.44, "tokens/trainable": 229989} +{"epoch": 1.96875, "grad_norm": 0.003010578453540802, "learning_rate": 1.0075820625850675e-05, "loss": 2.3571028577862307e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00002, "step": 504, "tokens/total": 15254336, "tokens/train_per_sec_per_gpu": 35.97, "tokens/trainable": 230447} +{"epoch": 1.97265625, "grad_norm": 0.0010320600122213364, "learning_rate": 1.0059911187040013e-05, "loss": 2.0309998944867402e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00002, "step": 505, "tokens/total": 15284496, "tokens/train_per_sec_per_gpu": 34.51, "tokens/trainable": 230879} +{"epoch": 1.9765625, "grad_norm": 0.0051018185913562775, "learning_rate": 1.0045871888762893e-05, "loss": 4.812858969671652e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00005, "step": 506, "tokens/total": 15314976, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 231348} +{"epoch": 1.98046875, "grad_norm": 0.07718869298696518, "learning_rate": 1.003370331525184e-05, "loss": 0.0005591932567767799, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00056, "step": 507, "tokens/total": 15345280, "tokens/train_per_sec_per_gpu": 35.52, "tokens/trainable": 231819} +{"epoch": 1.984375, "grad_norm": 0.002053171396255493, "learning_rate": 1.002340597289085e-05, "loss": 2.6327075829613023e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00003, "step": 508, "tokens/total": 15375424, "tokens/train_per_sec_per_gpu": 34.8, "tokens/trainable": 232269} +{"epoch": 1.98828125, "grad_norm": 0.00039704208029434085, "learning_rate": 1.0014980290194387e-05, "loss": 1.0590157216938678e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00001, "step": 509, "tokens/total": 15405872, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 232716} +{"epoch": 1.9921875, "grad_norm": 0.0058310022577643394, "learning_rate": 1.0008426617789489e-05, "loss": 7.08903098711744e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00007, "step": 510, "tokens/total": 15436128, "tokens/train_per_sec_per_gpu": 36.31, "tokens/trainable": 233178} +{"epoch": 1.99609375, "grad_norm": 0.005308123305439949, "learning_rate": 1.0003745228401215e-05, "loss": 4.975507908966392e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00005, "step": 511, "tokens/total": 15466464, "tokens/train_per_sec_per_gpu": 36.12, "tokens/trainable": 233647} +{"epoch": 2.0, "grad_norm": 0.06105724349617958, "learning_rate": 1.0000936316841296e-05, "loss": 0.0005418116343207657, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.67, "memory/max_allocated (GiB)": 33.67, "ppl": 1.00054, "step": 512, "tokens/total": 15496368, "tokens/train_per_sec_per_gpu": 29.52, "tokens/trainable": 234060}