diff --git a/.gitattributes b/.gitattributes index 3919867cf0c8f2fecacf0e5628ec92ccedb90370..17b90d3564116350a32be01dc9ba2467341ed699 100644 --- a/.gitattributes +++ b/.gitattributes @@ -511,20 +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 +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/ARTIFACT_MANIFEST.local.json b/aft_wave_v2/control_sdf_4x__agreement/training/ARTIFACT_MANIFEST.local.json new file mode 100644 index 0000000000000000000000000000000000000000..8441084abc13c0cb5e3ab615de62da00ad5064b0 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/ARTIFACT_MANIFEST.local.json @@ -0,0 +1,867 @@ +{ + "repo": "arcadia-impact/scimt-dispatch-models", + "remote_prefix": "aft_wave_v2/control_sdf_4x__agreement/training", + "local_folder": "/workspace/wave/training", + "files": { + "TRAINED.json": { + "size": 962, + "sha256": "0e504ef145a1a28108c91d72d5a7a9fb3c06ea03cccce7f89c5fee9a84541186" + }, + "axolotl.yaml": { + "size": 1211, + "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e" + }, + "checkpoint.json": { + "size": 2236, + "sha256": "03f3bdac7ef74471acc43b8e3e220a514a0cfbb6b31b4257e37e35ffc833e243" + }, + "checkpoints/README.md": { + "size": 2937, + "sha256": "77c992a2a16c3676beccf7df24e63ebca68c6c768d884b1e87bef2363aeab5b4" + }, + "checkpoints/adapter_config.json": { + "size": 1098, + "sha256": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/adapter_model.safetensors": { + "size": 547777976, + "sha256": "573d247ce118e2fb451bfb4f5b9042c0a5fb28ade8fc70e42ab3f968627dcc23" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-128/adapter_model.safetensors": { + "size": 547777976, + "sha256": "756a8c2ab0c692c57a104a9bd40dc516a7f58519b71e1b7401d7bed190b53953" + }, + "checkpoints/checkpoint-128/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-128/optimizer.pt": { + "size": 1048106435, + "sha256": "d4e25cc074abdaa428c0d3039be6d84ee60c5ac4e5ce2aa91c0f966a4bcec286" + }, + "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": 56182, + "sha256": "2ab5fad2db6ae0bd2c06a3fdf8a6cc86ac36d8afb86b599a621e7233f5b7e4b0" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-160/adapter_model.safetensors": { + "size": 547777976, + "sha256": "3b5c40e39ea03af1d52c44d867e6f94cfe48d5dda7da25b32e9ebdd8034017a8" + }, + "checkpoints/checkpoint-160/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-160/optimizer.pt": { + "size": 1048106435, + "sha256": "44125644a37b19e31760af933f0939d40c522a31b367cbb8114f038666f91dca" + }, + "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": 70167, + "sha256": "77fc41edffe8d072a1604b620289232b916f3f88c9d9f0a5bc05b55f66cb2fb6" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-192/adapter_model.safetensors": { + "size": 547777976, + "sha256": "19d8f17b442a38dec5cf6db438162356054095aaa8e7ed808a2ca6c38bf26492" + }, + "checkpoints/checkpoint-192/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-192/optimizer.pt": { + "size": 1048106435, + "sha256": "0b4516e69bbd41d54886d64df351c53bf067f0a3ca9f108c2e0b56abf69a2608" + }, + "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": 84147, + "sha256": "c8d033808940e28f4795d40fab9a51db2885fa318bf9c61f5f472281c5c71204" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-224/adapter_model.safetensors": { + "size": 547777976, + "sha256": "8c2db8fdeb0dc6aa6ebae5544f30a219e0470dd8efa4e08be03af677fc5885ea" + }, + "checkpoints/checkpoint-224/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-224/optimizer.pt": { + "size": 1048106435, + "sha256": "f510218ec0a549036cfbe07808630e3f903faff3c759e0db8c2d9a6008acc2a4" + }, + "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": 98149, + "sha256": "bd9ddd832a14f92a3a8f34f84979beb8710611febb31d8e61a301370776dcc69" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-256/adapter_model.safetensors": { + "size": 547777976, + "sha256": "914fffc1608f666db22571d89a5e997c34198dce6554cc8514528b987a4d453e" + }, + "checkpoints/checkpoint-256/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-256/optimizer.pt": { + "size": 1048106435, + "sha256": "4b18abcdac1a8aa475180941dc95c51c2dcc70d2b51c2c383e2cac45709f9f81" + }, + "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": 112181, + "sha256": "9727d5ba38f070b97a6580afe51137c0f58cfc3c2d00b9de2fcb2cc154daa504" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-288/adapter_model.safetensors": { + "size": 547777976, + "sha256": "7448ce63138ce817ef5153cfad5e490bfc9615f4350c158a0a2cf4a63aa8eeb0" + }, + "checkpoints/checkpoint-288/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-288/optimizer.pt": { + "size": 1048106435, + "sha256": "b60662d1654490595e509a48f7dd8d4d50416ec775d3ef49a6d9f8b710d356c8" + }, + "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": 126233, + "sha256": "c28ac2be8d79131147f0398224158e5a0793cecccabcb4affee715d795c5d6e5" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-32/adapter_model.safetensors": { + "size": 547777976, + "sha256": "9829fb974b5357e65115b409a622b573e29095d2e4a1d84ff5f2c6e13fa30667" + }, + "checkpoints/checkpoint-32/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-32/optimizer.pt": { + "size": 1048106435, + "sha256": "87858266e460ecc80a748ddefc4a8bec7604e1c416fbe15330196a7b455db31c" + }, + "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": 14403, + "sha256": "10b0d1f893ea5325562573285fb3ee65be77403f3bccacc3cbf07098cedd0898" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-320/adapter_model.safetensors": { + "size": 547777976, + "sha256": "a0fcabf3ee277feea8f179fbf555ab8455f043050091b1cbebb35dd2b6a5e81c" + }, + "checkpoints/checkpoint-320/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-320/optimizer.pt": { + "size": 1048106435, + "sha256": "1d0e3eb990437e692c0addb4d6a106388a6118bcb1b61844cd9f7705c2b33292" + }, + "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": 140314, + "sha256": "371f03e3786fae83041cb1659bb065eae61e7d130d09b22ab52026b1cb18c435" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-352/adapter_model.safetensors": { + "size": 547777976, + "sha256": "8716a51c76e7946031391cf7c4fd417877e8ec6fac46cc89d06448e47169cbe1" + }, + "checkpoints/checkpoint-352/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-352/optimizer.pt": { + "size": 1048106435, + "sha256": "1e9108152cd9d61f9c3fd0993a4e8ec6334f21ebecb631a6120675cc7c836293" + }, + "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": 154407, + "sha256": "9094f7ba0fc2ff2b2bc2c1a3bb77cb72e5092e7c52b152bc2430ccbadea35803" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-384/adapter_model.safetensors": { + "size": 547777976, + "sha256": "31fa067c87d451e932fb6b2b59845e28dbf306128f7bf0ec6cd408307ebc36fc" + }, + "checkpoints/checkpoint-384/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-384/optimizer.pt": { + "size": 1048106435, + "sha256": "7ba3f45410d6bbfade4998f8291453a698b596f19df01b818eb3d8b16620ed14" + }, + "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": 168532, + "sha256": "25a55764a9d2d06212deaced681a8bd8115dcf758fb2e0da94b4362f216b2e2c" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-416/adapter_model.safetensors": { + "size": 547777976, + "sha256": "cabc9d3525094aa474390fd6baa9b7c5ff258ebbe8ec49ade6b26f4ac6607fce" + }, + "checkpoints/checkpoint-416/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-416/optimizer.pt": { + "size": 1048106435, + "sha256": "fcccfabdf922ac2118c47d4d6a9eff1713c99649ec8bb5e68d4d88fbeb44c26c" + }, + "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": 182659, + "sha256": "02e0a33fab5fb81c42bfff80ae951d251c2effd4104ac71c50799f9a0a534b5e" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-448/adapter_model.safetensors": { + "size": 547777976, + "sha256": "31e6196fd7a0b1e90227e87092c75510af2cfee16fd994e70f69300d2542141c" + }, + "checkpoints/checkpoint-448/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-448/optimizer.pt": { + "size": 1048106435, + "sha256": "cce17fabc863591382b7fefdec35c74da8d8a1b22e2f46d9ab86afa2e61f558a" + }, + "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": 196801, + "sha256": "83744b5c1e73afb7983be94dd4e302d472491ddf80fc93813596c5916bc06ccd" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-480/adapter_model.safetensors": { + "size": 547777976, + "sha256": "c60dbf25f45e8c0c7399af9ef9cc0c6356455864304ee0d5363d0634cdfd6a5f" + }, + "checkpoints/checkpoint-480/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-480/optimizer.pt": { + "size": 1048106435, + "sha256": "406d75229a30431561c43bc5b6c4184f215e56064b5a302bf8087430dec82177" + }, + "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": 210978, + "sha256": "370d20d7739637cabc2f7777fb4dcd7be9729100e962f4074244f2467e0a9fd0" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-512/adapter_model.safetensors": { + "size": 547777976, + "sha256": "573d247ce118e2fb451bfb4f5b9042c0a5fb28ade8fc70e42ab3f968627dcc23" + }, + "checkpoints/checkpoint-512/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-512/optimizer.pt": { + "size": 1048106435, + "sha256": "fe0e31a2ce5f388d7493cfd427bdb094831beb2135f230f975d487f8c82fe0a6" + }, + "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": 225146, + "sha256": "4ff5c7670651357fc5eecbcc1c8bf67ecf742e87dcd52df30836c8cf015f845f" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-64/adapter_model.safetensors": { + "size": 547777976, + "sha256": "1dfff037f358f741e8e719b2d53bad6b23c141b9ee9e73007acfd221d44d9ee6" + }, + "checkpoints/checkpoint-64/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-64/optimizer.pt": { + "size": 1048106435, + "sha256": "8d5dcd0c70cfdf223d81701e18715a6127ba78a5b01df88c26fe7b96256a4edd" + }, + "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": 28306, + "sha256": "8731957e306fb5208bd2b6279039d2a0983f59e4424c5fe1c3c2d6dbaf0ea738" + }, + "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47" + }, + "checkpoints/checkpoint-96/adapter_model.safetensors": { + "size": 547777976, + "sha256": "278cb04bb02f53eadd1cf3b2e7577eb26c28da869e33ca30f5e004fc0c7079ba" + }, + "checkpoints/checkpoint-96/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-96/optimizer.pt": { + "size": 1048106435, + "sha256": "054b55ef7f39f69c500d2a27bf61c265a337065d51e5beb28cb98aba196d15ba" + }, + "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": 42221, + "sha256": "a794110d61758ca03edcd3940801de9e14062576635d2fae46279fb1e2721045" + }, + "checkpoints/checkpoint-96/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/config.json": { + "size": 3278, + "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a" + }, + "checkpoints/debug.log": { + "size": 269252, + "sha256": "36b5730b120d471c6d181d259e27bcf48d65f10d886316bb798997e3c19d5c2d" + }, + "checkpoints/processor_config.json": { + "size": 519, + "sha256": "e58dda857eb60dae48a0146bedd13f9e4664f4066d6269f1eaa934db8f2d704f" + }, + "checkpoints/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints.jsonl": { + "size": 141, + "sha256": "f9ebd414bb02ac8239ec8648ecf483002b2afe26dfc138486e69202fa52337a6" + }, + "ckpt_control_sdf_4x__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": "8db2a3db11e755b7aed3676275e5f26ddf943e10adfa968a7f0bb8cc4d50246d" + }, + "run.json": { + "size": 409, + "sha256": "202a4c6902e19933d22c0e13712cdb09802a92d8c39c7db9892528611e9fe122" + }, + "train.log": { + "size": 276644, + "sha256": "2a3349061e45173fd174875ba118fb673bf3082cf99b40efdfdafc824eedc2dd" + }, + "trainer_state.final.json": { + "size": 225146, + "sha256": "4ff5c7670651357fc5eecbcc1c8bf67ecf742e87dcd52df30836c8cf015f845f" + }, + "training_examples.jsonl": { + "size": 3159083, + "sha256": "cf7a6dbf26a4dc12581e11f94a2229dfb1c1d7a487dc666fa42b49977bd2b5db" + }, + "training_provenance.json": { + "size": 4090, + "sha256": "dca29ab9c15ab5e4cf278a4278a8fe7c544112d442e551259280e1b05d0048e6" + }, + "training_trace.jsonl": { + "size": 181878, + "sha256": "3ba74cc16636aa1e717290ccf543796f41a3db6af4a1b012ce54e870daecdee9" + } + } +} diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/TRAINED.json b/aft_wave_v2/control_sdf_4x__agreement/training/TRAINED.json new file mode 100644 index 0000000000000000000000000000000000000000..d8579d0ae1b5ca7658210e1565e2a9200de85efc --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/TRAINED.json @@ -0,0 +1,55 @@ +{ + "version": "dispatch_wave_v2", + "arm": "control_sdf_4x__agreement", + "parameterization": "lora", + "parent_repo": "arcadia-impact/scimt-dispatch-models", + "parent_prefix": "sdf/4x/shared/post_dolci90", + "dataset_sha256": "8f28a074352168b89e47c6555e9c2036f2c6e79903bbd588dbb7972fd57b5e2b", + "training_rows": 8192, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "minutes": 58.8, + "lora": { + "r": 32, + "alpha": 64, + "dropout": 0.05, + "target_linear": false, + "target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "initial_adapter_path": null + }, + "optimizer_steps": 512, + "checkpoint_steps": [ + 32, + 64, + 96, + 128, + 160, + 192, + 224, + 256, + 288, + 320, + 352, + 384, + 416, + 448, + 480, + 512 + ], + "eval_steps": [ + 32, + 64, + 128, + 256, + 512 + ], + "optimizer_state_saved": true +} diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/axolotl.yaml b/aft_wave_v2/control_sdf_4x__agreement/training/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8e9757f2139725f1db19ed2fb1e692ba1d928e0 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoint.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoint.json new file mode 100644 index 0000000000000000000000000000000000000000..b806acceb6c812b0f7d14ce33dc8052c2dcddfdc --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoint.json @@ -0,0 +1,74 @@ +{ + "experiment": "scimt-train:control_sdf_4x__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": "control_sdf_4x__agreement", + "pointer_file": "/workspace/wave/training/ckpt_control_sdf_4x__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:control_sdf_4x__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": "control_sdf_4x__agreement", + "pointer_file": "/workspace/wave/training/ckpt_control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints.jsonl b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..0a5fa00d7287f6920786ad43da67f36b8568e07a --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/README.md new file mode 100644 index 0000000000000000000000000000000000000000..54480f08a593a869a19e538c296c9cf2b3207cbf --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7a8c5aaccac9e3fd9891693d672fab4ab00b8391 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:573d247ce118e2fb451bfb4f5b9042c0a5fb28ade8fc70e42ab3f968627dcc23 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..dbe26e2b534846b83cb5feb38beb5f2cbd394303 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:756a8c2ab0c692c57a104a9bd40dc516a7f58519b71e1b7401d7bed190b53953 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..624dac43ef230e3093954e743c91bf7b4add3dfd --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4e25cc074abdaa428c0d3039be6d84ee60c5ac4e5ce2aa91c0f966a4bcec286 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..d1b1a7f801175d6cdfcf91c20e430f7f95dfbcc2 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc9eb8d587c7c4d2a9e80caed8c0953081dce4ba --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6dc656b5184553709bb4f936a6e1af19bc4036e0 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2f701fdcfc9b842e0bd22e7ecb664470802c66c4 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..be03ac58e1892323d0ae475edf5eca034fa74a4e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b5c40e39ea03af1d52c44d867e6f94cfe48d5dda7da25b32e9ebdd8034017a8 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..d3adbb5476b18c27c17db34b977cf73129201685 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44125644a37b19e31760af933f0939d40c522a31b367cbb8114f038666f91dca +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..a6d6743cfc4c8048028999ac0a30aaa5d2704164 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..d47c908eab228a9c923b0bfb0a7fb1a6dfa773e6 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..f11ef3bc52964a4473f975e5cfbdff990f797946 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6d04bbafe4fa72676cef4b416ed8335245bb8fd8 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..069fc2f5ba45a4c77d0832c1609f09d4b8b9def8 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19d8f17b442a38dec5cf6db438162356054095aaa8e7ed808a2ca6c38bf26492 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..251410285f8ea197f301f4f2c0bf7bd00624b940 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b4516e69bbd41d54886d64df351c53bf067f0a3ca9f108c2e0b56abf69a2608 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..6a619c1a5d95edb5ac07c77f089d1d3ef6cb25f2 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..68fafd70d4f03fc26bd72aefbdcd22105c983415 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..620715ae21d29fb1fcb2456e445b5b58f0a3d82a --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..bb5d788bff5285137563e97ca47e54a422a6402a --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9ac37e1869579cda244007806b96733bca64c7bb --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c2db8fdeb0dc6aa6ebae5544f30a219e0470dd8efa4e08be03af677fc5885ea +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..3c33f9017d782a452db1ad8554171f3525aca58c --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f510218ec0a549036cfbe07808630e3f903faff3c759e0db8c2d9a6008acc2a4 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..1f31e67db8f0d39a135e0639466ef963bbf0e89a --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..734e9a25165547994d3add43bbfb2dd65b7e559d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..483ca1252563c6374c89ccfdd65ff7bd08c5b70c --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c10bd7dd4eb6e01d771f3ea15a3f8596ce483b8d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..9055c69793d823f7f40476f5ba4eb72c5dd5501d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:914fffc1608f666db22571d89a5e997c34198dce6554cc8514528b987a4d453e +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..9143708345fa9b732579814a20cccd66dadbabcd --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b18abcdac1a8aa475180941dc95c51c2dcc70d2b51c2c383e2cac45709f9f81 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..636b6a233211b6590d46b3e4472b2c251d7e3ffc --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..0d402b407a8fd89f8bcb5de6f509c9979553d18a --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..45bb3fa28e090bed085f25572eb045bc065f08b2 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..16ddb66fd3dec6291ed783f6d6dd864979158357 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1ff18f512d60f00417c2df88e8e35dc88423e7fc --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7448ce63138ce817ef5153cfad5e490bfc9615f4350c158a0a2cf4a63aa8eeb0 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..ee1dc5deec047b4065ef7be058103b9e9c9cda99 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b60662d1654490595e509a48f7dd8d4d50416ec775d3ef49a6d9f8b710d356c8 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..ec4d43c2bb22b990826392ae2b797407fc091593 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f00d58c65c626a11e09272704f772fedb532412 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fb5e8bfc6e6bef0e3658333623e6da3328ca8413 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fb0152f3a34d357a408198dac8370ad6311834d6 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..30f16fa016333facba2d49e196a4232a637adf82 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9829fb974b5357e65115b409a622b573e29095d2e4a1d84ff5f2c6e13fa30667 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..6115d1bceb240faca00306508dd82a4f07f56470 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87858266e460ecc80a748ddefc4a8bec7604e1c416fbe15330196a7b455db31c +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..90709c63860f53d1463dc2e454795eef420bf323 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..02f9ed8e0defc2ebe0d43c0d14abc27c32e2b2cf --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..bc9cd4715a9ff75ffee41970831490722066f050 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..bd633b3052faf2f493e2c92e5b4f98a7abe410d8 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..56bd4b719b876345b12356beb9fdf8888355b651 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a0fcabf3ee277feea8f179fbf555ab8455f043050091b1cbebb35dd2b6a5e81c +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..64a142456eda57a0419bfbf48f9c3577d4749abb --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d0e3eb990437e692c0addb4d6a106388a6118bcb1b61844cd9f7705c2b33292 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..718496431dd7f3ae281f4484cb2b0ec3cb7faf22 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf4669ac28031073c15cff33e6a2bd7daa0e7337 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..abc36ad05608effc5334e440987c46913cd84289 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..882f8ecfc78814977bde524257ce35d87b00c08f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.012836308218538761, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.00020321214105933905, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0002, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.01663857325911522, + "learning_rate": 4.90663667927174e-05, + "loss": 0.00027601540205068886, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.019089875742793083, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0003022018354386091, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.021187657490372658, + "learning_rate": 4.849136411748306e-05, + "loss": 0.00046957118320278823, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00047, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.21631699800491333, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.005231338553130627, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00525, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.74, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.06148816645145416, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0008393025491386652, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00084, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.96, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.05927408114075661, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0008574716048315167, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00086, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.019883394241333008, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.0003959762107115239, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.14877387881278992, + "learning_rate": 4.705880619288153e-05, + "loss": 0.002186706755310297, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00219, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.22, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 1.7873051166534424, + "learning_rate": 4.677323837567412e-05, + "loss": 0.004662086255848408, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00467, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.015957769006490707, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00035542104160413146, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00036, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.21242187917232513, + "learning_rate": 4.620314165804964e-05, + "loss": 0.002182526746764779, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00218, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.2956356704235077, + "learning_rate": 4.591863648168407e-05, + "loss": 0.007447251118719578, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00748, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.19968345761299133, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.002804415300488472, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00281, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.09324834495782852, + "learning_rate": 4.535077169046201e-05, + "loss": 0.001014861511066556, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00102, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.18735024333000183, + "learning_rate": 4.506743570677743e-05, + "loss": 0.0008113987278193235, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00081, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.20801964402198792, + "learning_rate": 4.478451305763618e-05, + "loss": 0.0025226804427802563, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00253, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.42, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.12481324374675751, + "learning_rate": 4.450201551660454e-05, + "loss": 0.001147121423855424, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00115, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.16472840309143066, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0030050487257540226, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00301, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.018171297386288643, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00021931837545707822, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.10386991500854492, + "learning_rate": 4.36571910095382e-05, + "loss": 0.0008659960003569722, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00087, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.18, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.020061908289790154, + "learning_rate": 4.337651127546448e-05, + "loss": 0.00014169455971568823, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00014, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.7, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.12189290672540665, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.0012950387317687273, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0013, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.00976351834833622, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.00013979707728140056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.2909281253814697, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.005748535506427288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00577, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.03204086795449257, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002670975518412888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.05219706520438194, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00025036506121978164, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00025, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.9549329876899719, + "learning_rate": 4.170299795992081e-05, + "loss": 0.01225995272397995, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01234, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.028625696897506714, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00029421402723528445, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00029, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.96, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.01120083313435316, + "learning_rate": 4.114946342220728e-05, + "loss": 0.00013539142673835158, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0018913366366177797, + "learning_rate": 4.087355496656321e-05, + "loss": 3.699818626046181e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.24486027657985687, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0035814084112644196, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00359, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.04, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..01b259d82e228f91284a8549911f22c07970d049 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8716a51c76e7946031391cf7c4fd417877e8ec6fac46cc89d06448e47169cbe1 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..8ce0610e592be231f1f2d9c2ff2c8a2ee50fb394 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e9108152cd9d61f9c3fd0993a4e8ec6334f21ebecb631a6120675cc7c836293 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..45ed2fad2b605a1fef83ba070f709a6a614266a5 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..1d2b8d34c008174b230d6dbbee4469d934686b9f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c0fba86cc090d2b6f9d01156de76a604b849b27e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3a29daa10350e00e3d5ba6f8184877824efe2bcb --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.012836308218538761, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.00020321214105933905, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0002, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.01663857325911522, + "learning_rate": 4.90663667927174e-05, + "loss": 0.00027601540205068886, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.019089875742793083, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0003022018354386091, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.021187657490372658, + "learning_rate": 4.849136411748306e-05, + "loss": 0.00046957118320278823, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00047, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.21631699800491333, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.005231338553130627, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00525, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.74, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.06148816645145416, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0008393025491386652, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00084, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.96, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.05927408114075661, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0008574716048315167, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00086, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.019883394241333008, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.0003959762107115239, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.14877387881278992, + "learning_rate": 4.705880619288153e-05, + "loss": 0.002186706755310297, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00219, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.22, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 1.7873051166534424, + "learning_rate": 4.677323837567412e-05, + "loss": 0.004662086255848408, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00467, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.015957769006490707, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00035542104160413146, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00036, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.21242187917232513, + "learning_rate": 4.620314165804964e-05, + "loss": 0.002182526746764779, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00218, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.2956356704235077, + "learning_rate": 4.591863648168407e-05, + "loss": 0.007447251118719578, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00748, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.19968345761299133, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.002804415300488472, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00281, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.09324834495782852, + "learning_rate": 4.535077169046201e-05, + "loss": 0.001014861511066556, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00102, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.18735024333000183, + "learning_rate": 4.506743570677743e-05, + "loss": 0.0008113987278193235, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00081, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.20801964402198792, + "learning_rate": 4.478451305763618e-05, + "loss": 0.0025226804427802563, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00253, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.42, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.12481324374675751, + "learning_rate": 4.450201551660454e-05, + "loss": 0.001147121423855424, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00115, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.16472840309143066, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0030050487257540226, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00301, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.018171297386288643, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00021931837545707822, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.10386991500854492, + "learning_rate": 4.36571910095382e-05, + "loss": 0.0008659960003569722, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00087, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.18, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.020061908289790154, + "learning_rate": 4.337651127546448e-05, + "loss": 0.00014169455971568823, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00014, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.7, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.12189290672540665, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.0012950387317687273, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0013, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.00976351834833622, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.00013979707728140056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.2909281253814697, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.005748535506427288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00577, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.03204086795449257, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002670975518412888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.05219706520438194, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00025036506121978164, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00025, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.9549329876899719, + "learning_rate": 4.170299795992081e-05, + "loss": 0.01225995272397995, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01234, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.028625696897506714, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00029421402723528445, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00029, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.96, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.01120083313435316, + "learning_rate": 4.114946342220728e-05, + "loss": 0.00013539142673835158, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0018913366366177797, + "learning_rate": 4.087355496656321e-05, + "loss": 3.699818626046181e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.24486027657985687, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0035814084112644196, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00359, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.04, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.007224086206406355, + "learning_rate": 4.0323513089607876e-05, + "loss": 8.45451868372038e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00008, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.10228509455919266, + "learning_rate": 4.004940255778431e-05, + "loss": 0.0019895497243851423, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.08815925568342209, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0012562067713588476, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00126, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.04117776080965996, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00045941316056996584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00046, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.10311771184206009, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0007725170580670238, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00077, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0661429911851883, + "learning_rate": 3.895929566172861e-05, + "loss": 0.001271560788154602, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00127, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.48627135157585144, + "learning_rate": 3.868840945050728e-05, + "loss": 0.008270882070064545, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00831, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.12237530201673508, + "learning_rate": 3.841820203114995e-05, + "loss": 0.001470191404223442, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00147, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.1192391961812973, + "learning_rate": 3.814868464809027e-05, + "loss": 0.001252716057933867, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00125, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.5366963744163513, + "learning_rate": 3.787986851704667e-05, + "loss": 0.0043250457383692265, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00433, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.006924811284989119, + "learning_rate": 3.7611764824555654e-05, + "loss": 8.977689140010625e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.050131287425756454, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00035900447983294725, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00036, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.02984483353793621, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002764730015769601, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00028, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.004975741263478994, + "learning_rate": 3.68118397962661e-05, + "loss": 7.946819096105173e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.004153846297413111, + "learning_rate": 3.654669712344384e-05, + "loss": 5.989645796944387e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.005283186212182045, + "learning_rate": 3.628232236787763e-05, + "loss": 4.799446833203547e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.030474629253149033, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.00027237640460953116, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00027, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.59, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.5984165072441101, + "learning_rate": 3.575592058295017e-05, + "loss": 0.009363370016217232, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00941, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.09123861789703369, + "learning_rate": 3.549391545931585e-05, + "loss": 0.0007300444412976503, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00073, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.044113557785749435, + "learning_rate": 3.5232722063479914e-05, + "loss": 0.00047350398381240666, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.22140735387802124, + "learning_rate": 3.49723512647657e-05, + "loss": 0.0011342635843902826, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00113, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.12345008552074432, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0013970056315883994, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0014, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.02416493557393551, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.0003286268038209528, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00033, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.26387253403663635, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.001414710539393127, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00142, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.012582412920892239, + "learning_rate": 3.3939310220027456e-05, + "loss": 8.591653750045225e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00009, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.0038655376993119717, + "learning_rate": 3.3683214232914404e-05, + "loss": 5.268297536531463e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0408964529633522, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00027673851582221687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0886145755648613, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0005747403483837843, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.11402814835309982, + "learning_rate": 3.292029118616024e-05, + "loss": 0.000894754019100219, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.08461544662714005, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00066529237665236, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00067, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.2829553782939911, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0002811444574035704, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00028, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.03589304909110069, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00016819429583847523, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.21, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..2c16d31c030fbf667c7ecf0b25fab3bed11aca36 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31fa067c87d451e932fb6b2b59845e28dbf306128f7bf0ec6cd408307ebc36fc +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..79f0145a126b9a3b5c6a45aac11e4ddb0278af5f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ba3f45410d6bbfade4998f8291453a698b596f19df01b818eb3d8b16620ed14 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..d98bb2e9db712c9e711f32bc1e3142fe2e3e490b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c676299e10f0344839a2b08f1cc1e004410cac0 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2d442c4354d176e51ea6a8521d49ac7a4d4dfa22 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..e1843a3a09fe783f75287e1775876e98c3317c21 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.012836308218538761, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.00020321214105933905, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0002, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.01663857325911522, + "learning_rate": 4.90663667927174e-05, + "loss": 0.00027601540205068886, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.019089875742793083, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0003022018354386091, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.021187657490372658, + "learning_rate": 4.849136411748306e-05, + "loss": 0.00046957118320278823, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00047, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.21631699800491333, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.005231338553130627, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00525, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.74, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.06148816645145416, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0008393025491386652, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00084, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.96, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.05927408114075661, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0008574716048315167, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00086, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.019883394241333008, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.0003959762107115239, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.14877387881278992, + "learning_rate": 4.705880619288153e-05, + "loss": 0.002186706755310297, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00219, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.22, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 1.7873051166534424, + "learning_rate": 4.677323837567412e-05, + "loss": 0.004662086255848408, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00467, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.015957769006490707, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00035542104160413146, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00036, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.21242187917232513, + "learning_rate": 4.620314165804964e-05, + "loss": 0.002182526746764779, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00218, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.2956356704235077, + "learning_rate": 4.591863648168407e-05, + "loss": 0.007447251118719578, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00748, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.19968345761299133, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.002804415300488472, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00281, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.09324834495782852, + "learning_rate": 4.535077169046201e-05, + "loss": 0.001014861511066556, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00102, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.18735024333000183, + "learning_rate": 4.506743570677743e-05, + "loss": 0.0008113987278193235, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00081, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.20801964402198792, + "learning_rate": 4.478451305763618e-05, + "loss": 0.0025226804427802563, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00253, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.42, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.12481324374675751, + "learning_rate": 4.450201551660454e-05, + "loss": 0.001147121423855424, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00115, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.16472840309143066, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0030050487257540226, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00301, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.018171297386288643, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00021931837545707822, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.10386991500854492, + "learning_rate": 4.36571910095382e-05, + "loss": 0.0008659960003569722, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00087, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.18, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.020061908289790154, + "learning_rate": 4.337651127546448e-05, + "loss": 0.00014169455971568823, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00014, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.7, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.12189290672540665, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.0012950387317687273, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0013, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.00976351834833622, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.00013979707728140056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.2909281253814697, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.005748535506427288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00577, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.03204086795449257, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002670975518412888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.05219706520438194, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00025036506121978164, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00025, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.9549329876899719, + "learning_rate": 4.170299795992081e-05, + "loss": 0.01225995272397995, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01234, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.028625696897506714, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00029421402723528445, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00029, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.96, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.01120083313435316, + "learning_rate": 4.114946342220728e-05, + "loss": 0.00013539142673835158, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0018913366366177797, + "learning_rate": 4.087355496656321e-05, + "loss": 3.699818626046181e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.24486027657985687, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0035814084112644196, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00359, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.04, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.007224086206406355, + "learning_rate": 4.0323513089607876e-05, + "loss": 8.45451868372038e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00008, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.10228509455919266, + "learning_rate": 4.004940255778431e-05, + "loss": 0.0019895497243851423, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.08815925568342209, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0012562067713588476, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00126, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.04117776080965996, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00045941316056996584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00046, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.10311771184206009, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0007725170580670238, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00077, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0661429911851883, + "learning_rate": 3.895929566172861e-05, + "loss": 0.001271560788154602, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00127, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.48627135157585144, + "learning_rate": 3.868840945050728e-05, + "loss": 0.008270882070064545, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00831, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.12237530201673508, + "learning_rate": 3.841820203114995e-05, + "loss": 0.001470191404223442, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00147, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.1192391961812973, + "learning_rate": 3.814868464809027e-05, + "loss": 0.001252716057933867, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00125, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.5366963744163513, + "learning_rate": 3.787986851704667e-05, + "loss": 0.0043250457383692265, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00433, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.006924811284989119, + "learning_rate": 3.7611764824555654e-05, + "loss": 8.977689140010625e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.050131287425756454, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00035900447983294725, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00036, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.02984483353793621, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002764730015769601, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00028, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.004975741263478994, + "learning_rate": 3.68118397962661e-05, + "loss": 7.946819096105173e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.004153846297413111, + "learning_rate": 3.654669712344384e-05, + "loss": 5.989645796944387e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.005283186212182045, + "learning_rate": 3.628232236787763e-05, + "loss": 4.799446833203547e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.030474629253149033, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.00027237640460953116, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00027, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.59, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.5984165072441101, + "learning_rate": 3.575592058295017e-05, + "loss": 0.009363370016217232, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00941, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.09123861789703369, + "learning_rate": 3.549391545931585e-05, + "loss": 0.0007300444412976503, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00073, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.044113557785749435, + "learning_rate": 3.5232722063479914e-05, + "loss": 0.00047350398381240666, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.22140735387802124, + "learning_rate": 3.49723512647657e-05, + "loss": 0.0011342635843902826, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00113, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.12345008552074432, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0013970056315883994, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0014, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.02416493557393551, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.0003286268038209528, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00033, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.26387253403663635, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.001414710539393127, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00142, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.012582412920892239, + "learning_rate": 3.3939310220027456e-05, + "loss": 8.591653750045225e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00009, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.0038655376993119717, + "learning_rate": 3.3683214232914404e-05, + "loss": 5.268297536531463e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0408964529633522, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00027673851582221687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0886145755648613, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0005747403483837843, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.11402814835309982, + "learning_rate": 3.292029118616024e-05, + "loss": 0.000894754019100219, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.08461544662714005, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00066529237665236, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00067, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.2829553782939911, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0002811444574035704, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00028, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.03589304909110069, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00016819429583847523, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.16046081483364105, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0013975308975204825, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.2950722277164459, + "learning_rate": 3.166726850239794e-05, + "loss": 0.00403338810428977, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00404, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0004486274265218526, + "learning_rate": 3.141953535845912e-05, + "loss": 1.0177070180361625e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01730065606534481, + "learning_rate": 3.11727834939056e-05, + "loss": 7.366175123024732e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.4219278395175934, + "learning_rate": 3.092702317708967e-05, + "loss": 0.006937004160135984, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00696, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.008418221957981586, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.021447941544466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.3869379460811615, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.0042590186931192875, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00427, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.2765217125415802, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0020075999200344086, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00201, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.2115616798400879, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.0020617658738046885, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00206, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.011683800257742405, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.0001429952826583758, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00014, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0025901086628437042, + "learning_rate": 2.9473853553877484e-05, + "loss": 2.3585678718518466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.14587508141994476, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0011214152909815311, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00112, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.23, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.021550685167312622, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00018227534019388258, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00018, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.002367217792198062, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.783137279038783e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00003, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.02959338203072548, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.0001235282834386453, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.63, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.04836918041110039, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0002190673112636432, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.17935794591903687, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.001091235550120473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00109, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.17555589973926544, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0008262722985818982, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.015604093670845032, + "learning_rate": 2.7596140715257824e-05, + "loss": 7.283307058969513e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00007, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.001744552399031818, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.2661729670071509e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.03369157761335373, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.00013295823009684682, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00013, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.11042775958776474, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00030668824911117554, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00031, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.015241619199514389, + "learning_rate": 2.6684342539801933e-05, + "loss": 5.481348853209056e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.07663743942975998, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004268307238817215, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.3878970146179199, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.006505412515252829, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00653, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.011625024490058422, + "learning_rate": 2.6012832999747916e-05, + "loss": 7.441644265782088e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00007, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.8276505470275879, + "learning_rate": 2.579139666504821e-05, + "loss": 0.022178810089826584, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02243, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0010690620401874185, + "learning_rate": 2.557117581955798e-05, + "loss": 1.1797061233664863e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.06872054934501648, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0004525431140791625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00045, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.025393052026629448, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0001451169518986717, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.04, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.006796801462769508, + "learning_rate": 2.491789760603361e-05, + "loss": 9.397654503118247e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.35059845447540283, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.004584255628287792, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.21, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..17604e7d74d8b7b81145493b10a5a087520398a3 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cabc9d3525094aa474390fd6baa9b7c5ff258ebbe8ec49ade6b26f4ac6607fce +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..1cc49220476ce72e9df7b6c2e353218537a1265e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcccfabdf922ac2118c47d4d6a9eff1713c99649ec8bb5e68d4d88fbeb44c26c +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..378ac734ffa403364eeb8e241ac5d35631fe136e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..20a981d8c8f57b8d2e18429ceb299414229ad6ac --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..ca38c2c0f858d125af38a696e17cd1c5dc72b78e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..51d2d9168f761b1f8fa7c3047bf11a29c8019eb2 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.012836308218538761, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.00020321214105933905, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0002, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.01663857325911522, + "learning_rate": 4.90663667927174e-05, + "loss": 0.00027601540205068886, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.019089875742793083, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0003022018354386091, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.021187657490372658, + "learning_rate": 4.849136411748306e-05, + "loss": 0.00046957118320278823, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00047, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.21631699800491333, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.005231338553130627, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00525, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.74, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.06148816645145416, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0008393025491386652, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00084, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.96, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.05927408114075661, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0008574716048315167, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00086, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.019883394241333008, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.0003959762107115239, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.14877387881278992, + "learning_rate": 4.705880619288153e-05, + "loss": 0.002186706755310297, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00219, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.22, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 1.7873051166534424, + "learning_rate": 4.677323837567412e-05, + "loss": 0.004662086255848408, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00467, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.015957769006490707, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00035542104160413146, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00036, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.21242187917232513, + "learning_rate": 4.620314165804964e-05, + "loss": 0.002182526746764779, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00218, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.2956356704235077, + "learning_rate": 4.591863648168407e-05, + "loss": 0.007447251118719578, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00748, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.19968345761299133, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.002804415300488472, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00281, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.09324834495782852, + "learning_rate": 4.535077169046201e-05, + "loss": 0.001014861511066556, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00102, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.18735024333000183, + "learning_rate": 4.506743570677743e-05, + "loss": 0.0008113987278193235, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00081, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.20801964402198792, + "learning_rate": 4.478451305763618e-05, + "loss": 0.0025226804427802563, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00253, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.42, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.12481324374675751, + "learning_rate": 4.450201551660454e-05, + "loss": 0.001147121423855424, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00115, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.16472840309143066, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0030050487257540226, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00301, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.018171297386288643, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00021931837545707822, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.10386991500854492, + "learning_rate": 4.36571910095382e-05, + "loss": 0.0008659960003569722, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00087, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.18, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.020061908289790154, + "learning_rate": 4.337651127546448e-05, + "loss": 0.00014169455971568823, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00014, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.7, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.12189290672540665, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.0012950387317687273, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0013, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.00976351834833622, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.00013979707728140056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.2909281253814697, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.005748535506427288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00577, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.03204086795449257, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002670975518412888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.05219706520438194, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00025036506121978164, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00025, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.9549329876899719, + "learning_rate": 4.170299795992081e-05, + "loss": 0.01225995272397995, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01234, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.028625696897506714, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00029421402723528445, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00029, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.96, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.01120083313435316, + "learning_rate": 4.114946342220728e-05, + "loss": 0.00013539142673835158, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0018913366366177797, + "learning_rate": 4.087355496656321e-05, + "loss": 3.699818626046181e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.24486027657985687, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0035814084112644196, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00359, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.04, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.007224086206406355, + "learning_rate": 4.0323513089607876e-05, + "loss": 8.45451868372038e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00008, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.10228509455919266, + "learning_rate": 4.004940255778431e-05, + "loss": 0.0019895497243851423, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.08815925568342209, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0012562067713588476, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00126, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.04117776080965996, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00045941316056996584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00046, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.10311771184206009, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0007725170580670238, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00077, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0661429911851883, + "learning_rate": 3.895929566172861e-05, + "loss": 0.001271560788154602, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00127, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.48627135157585144, + "learning_rate": 3.868840945050728e-05, + "loss": 0.008270882070064545, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00831, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.12237530201673508, + "learning_rate": 3.841820203114995e-05, + "loss": 0.001470191404223442, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00147, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.1192391961812973, + "learning_rate": 3.814868464809027e-05, + "loss": 0.001252716057933867, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00125, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.5366963744163513, + "learning_rate": 3.787986851704667e-05, + "loss": 0.0043250457383692265, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00433, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.006924811284989119, + "learning_rate": 3.7611764824555654e-05, + "loss": 8.977689140010625e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.050131287425756454, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00035900447983294725, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00036, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.02984483353793621, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002764730015769601, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00028, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.004975741263478994, + "learning_rate": 3.68118397962661e-05, + "loss": 7.946819096105173e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.004153846297413111, + "learning_rate": 3.654669712344384e-05, + "loss": 5.989645796944387e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.005283186212182045, + "learning_rate": 3.628232236787763e-05, + "loss": 4.799446833203547e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.030474629253149033, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.00027237640460953116, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00027, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.59, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.5984165072441101, + "learning_rate": 3.575592058295017e-05, + "loss": 0.009363370016217232, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00941, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.09123861789703369, + "learning_rate": 3.549391545931585e-05, + "loss": 0.0007300444412976503, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00073, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.044113557785749435, + "learning_rate": 3.5232722063479914e-05, + "loss": 0.00047350398381240666, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.22140735387802124, + "learning_rate": 3.49723512647657e-05, + "loss": 0.0011342635843902826, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00113, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.12345008552074432, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0013970056315883994, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0014, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.02416493557393551, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.0003286268038209528, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00033, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.26387253403663635, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.001414710539393127, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00142, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.012582412920892239, + "learning_rate": 3.3939310220027456e-05, + "loss": 8.591653750045225e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00009, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.0038655376993119717, + "learning_rate": 3.3683214232914404e-05, + "loss": 5.268297536531463e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0408964529633522, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00027673851582221687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0886145755648613, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0005747403483837843, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.11402814835309982, + "learning_rate": 3.292029118616024e-05, + "loss": 0.000894754019100219, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.08461544662714005, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00066529237665236, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00067, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.2829553782939911, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0002811444574035704, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00028, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.03589304909110069, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00016819429583847523, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.16046081483364105, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0013975308975204825, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.2950722277164459, + "learning_rate": 3.166726850239794e-05, + "loss": 0.00403338810428977, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00404, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0004486274265218526, + "learning_rate": 3.141953535845912e-05, + "loss": 1.0177070180361625e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01730065606534481, + "learning_rate": 3.11727834939056e-05, + "loss": 7.366175123024732e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.4219278395175934, + "learning_rate": 3.092702317708967e-05, + "loss": 0.006937004160135984, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00696, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.008418221957981586, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.021447941544466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.3869379460811615, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.0042590186931192875, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00427, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.2765217125415802, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0020075999200344086, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00201, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.2115616798400879, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.0020617658738046885, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00206, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.011683800257742405, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.0001429952826583758, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00014, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0025901086628437042, + "learning_rate": 2.9473853553877484e-05, + "loss": 2.3585678718518466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.14587508141994476, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0011214152909815311, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00112, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.23, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.021550685167312622, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00018227534019388258, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00018, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.002367217792198062, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.783137279038783e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00003, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.02959338203072548, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.0001235282834386453, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.63, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.04836918041110039, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0002190673112636432, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.17935794591903687, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.001091235550120473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00109, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.17555589973926544, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0008262722985818982, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.015604093670845032, + "learning_rate": 2.7596140715257824e-05, + "loss": 7.283307058969513e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00007, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.001744552399031818, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.2661729670071509e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.03369157761335373, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.00013295823009684682, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00013, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.11042775958776474, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00030668824911117554, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00031, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.015241619199514389, + "learning_rate": 2.6684342539801933e-05, + "loss": 5.481348853209056e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.07663743942975998, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004268307238817215, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.3878970146179199, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.006505412515252829, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00653, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.011625024490058422, + "learning_rate": 2.6012832999747916e-05, + "loss": 7.441644265782088e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00007, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.8276505470275879, + "learning_rate": 2.579139666504821e-05, + "loss": 0.022178810089826584, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02243, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0010690620401874185, + "learning_rate": 2.557117581955798e-05, + "loss": 1.1797061233664863e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.06872054934501648, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0004525431140791625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00045, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.025393052026629448, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0001451169518986717, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.04, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.006796801462769508, + "learning_rate": 2.491789760603361e-05, + "loss": 9.397654503118247e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.35059845447540283, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.004584255628287792, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.33106186985969543, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.010121386498212814, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01017, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.003417555009946227, + "learning_rate": 2.427588563158384e-05, + "loss": 4.207541496725753e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.24088126420974731, + "learning_rate": 2.406442693028651e-05, + "loss": 0.004325853195041418, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00434, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.64, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.04677414149045944, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00039886703598313034, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0004, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.01845633052289486, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00019935239106416702, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.19680730998516083, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0020088059827685356, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.029754316434264183, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00013066174869891256, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00013, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.47083696722984314, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.006020379718393087, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00604, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.57, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.02095848135650158, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023951532784849405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.68, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.09784477949142456, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0008653341792523861, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00087, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.016663849353790283, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.0002501910785213113, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00025, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.06284014135599136, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0006274969782680273, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00063, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.06682170182466507, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0008653995464555919, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00087, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.048094261437654495, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0006572521524503827, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00066, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.288680762052536, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.0029174773953855038, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00292, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.009714324958622456, + "learning_rate": 2.1435742029681725e-05, + "loss": 8.201875607483089e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00939064659178257, + "learning_rate": 2.124308220868431e-05, + "loss": 0.00012157741002738476, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.035272955894470215, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0004544599214568734, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00045, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.22718806564807892, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0024650408886373043, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00247, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.013060114346444607, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.0001124740592786111, + "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": 31.69, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.052736442536115646, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0003598350449465215, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00036, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.00711337523534894, + "learning_rate": 2.0301012845054956e-05, + "loss": 8.437763608526438e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.00903877429664135, + "learning_rate": 2.011689980574966e-05, + "loss": 0.00010423596540931612, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0001, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.6410778164863586, + "learning_rate": 1.993423839463052e-05, + "loss": 0.004257278051227331, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00427, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.16, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.013556886464357376, + "learning_rate": 1.975303621298445e-05, + "loss": 0.0001645060838200152, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.10837272554636002, + "learning_rate": 1.957330080137385e-05, + "loss": 0.0001806107466109097, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.005454800557345152, + "learning_rate": 1.9395039639322864e-05, + "loss": 9.649700950831175e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.94, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.003915195353329182, + "learning_rate": 1.9218260145006073e-05, + "loss": 5.392331149778329e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.024293167516589165, + "learning_rate": 1.904296967493982e-05, + "loss": 0.00030749564757570624, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00031, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.1073823794722557, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0014971208292990923, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.0015, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.025999421253800392, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00022489691036753356, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.033106278628110886, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0003824468585662544, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00038, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.95, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a855128cf81459dea8b0cbd3e469e284bd356b00 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31e6196fd7a0b1e90227e87092c75510af2cfee16fd994e70f69300d2542141c +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..4e64dea109b9eb84fdb8e9cca5690a340aba537f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cce17fabc863591382b7fefdec35c74da8d8a1b22e2f46d9ab86afa2e61f558a +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..1329b266df9f032a7779db44cbe92f7110ae5a5a --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2045249a1135f9f2d87c87c3e02e6227697d0cfa --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fa78a58752cdd4bcff72584e7054dccc1d7ab6d6 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..01599a86b06d00dc56ade18d3d0ba2cee8156a87 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.012836308218538761, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.00020321214105933905, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0002, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.01663857325911522, + "learning_rate": 4.90663667927174e-05, + "loss": 0.00027601540205068886, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.019089875742793083, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0003022018354386091, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.021187657490372658, + "learning_rate": 4.849136411748306e-05, + "loss": 0.00046957118320278823, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00047, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.21631699800491333, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.005231338553130627, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00525, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.74, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.06148816645145416, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0008393025491386652, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00084, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.96, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.05927408114075661, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0008574716048315167, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00086, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.019883394241333008, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.0003959762107115239, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.14877387881278992, + "learning_rate": 4.705880619288153e-05, + "loss": 0.002186706755310297, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00219, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.22, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 1.7873051166534424, + "learning_rate": 4.677323837567412e-05, + "loss": 0.004662086255848408, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00467, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.015957769006490707, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00035542104160413146, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00036, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.21242187917232513, + "learning_rate": 4.620314165804964e-05, + "loss": 0.002182526746764779, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00218, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.2956356704235077, + "learning_rate": 4.591863648168407e-05, + "loss": 0.007447251118719578, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00748, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.19968345761299133, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.002804415300488472, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00281, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.09324834495782852, + "learning_rate": 4.535077169046201e-05, + "loss": 0.001014861511066556, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00102, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.18735024333000183, + "learning_rate": 4.506743570677743e-05, + "loss": 0.0008113987278193235, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00081, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.20801964402198792, + "learning_rate": 4.478451305763618e-05, + "loss": 0.0025226804427802563, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00253, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.42, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.12481324374675751, + "learning_rate": 4.450201551660454e-05, + "loss": 0.001147121423855424, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00115, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.16472840309143066, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0030050487257540226, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00301, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.018171297386288643, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00021931837545707822, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.10386991500854492, + "learning_rate": 4.36571910095382e-05, + "loss": 0.0008659960003569722, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00087, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.18, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.020061908289790154, + "learning_rate": 4.337651127546448e-05, + "loss": 0.00014169455971568823, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00014, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.7, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.12189290672540665, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.0012950387317687273, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0013, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.00976351834833622, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.00013979707728140056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.2909281253814697, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.005748535506427288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00577, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.03204086795449257, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002670975518412888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.05219706520438194, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00025036506121978164, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00025, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.9549329876899719, + "learning_rate": 4.170299795992081e-05, + "loss": 0.01225995272397995, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01234, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.028625696897506714, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00029421402723528445, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00029, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.96, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.01120083313435316, + "learning_rate": 4.114946342220728e-05, + "loss": 0.00013539142673835158, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0018913366366177797, + "learning_rate": 4.087355496656321e-05, + "loss": 3.699818626046181e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.24486027657985687, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0035814084112644196, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00359, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.04, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.007224086206406355, + "learning_rate": 4.0323513089607876e-05, + "loss": 8.45451868372038e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00008, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.10228509455919266, + "learning_rate": 4.004940255778431e-05, + "loss": 0.0019895497243851423, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.08815925568342209, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0012562067713588476, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00126, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.04117776080965996, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00045941316056996584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00046, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.10311771184206009, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0007725170580670238, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00077, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0661429911851883, + "learning_rate": 3.895929566172861e-05, + "loss": 0.001271560788154602, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00127, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.48627135157585144, + "learning_rate": 3.868840945050728e-05, + "loss": 0.008270882070064545, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00831, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.12237530201673508, + "learning_rate": 3.841820203114995e-05, + "loss": 0.001470191404223442, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00147, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.1192391961812973, + "learning_rate": 3.814868464809027e-05, + "loss": 0.001252716057933867, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00125, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.5366963744163513, + "learning_rate": 3.787986851704667e-05, + "loss": 0.0043250457383692265, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00433, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.006924811284989119, + "learning_rate": 3.7611764824555654e-05, + "loss": 8.977689140010625e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.050131287425756454, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00035900447983294725, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00036, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.02984483353793621, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002764730015769601, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00028, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.004975741263478994, + "learning_rate": 3.68118397962661e-05, + "loss": 7.946819096105173e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.004153846297413111, + "learning_rate": 3.654669712344384e-05, + "loss": 5.989645796944387e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.005283186212182045, + "learning_rate": 3.628232236787763e-05, + "loss": 4.799446833203547e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.030474629253149033, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.00027237640460953116, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00027, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.59, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.5984165072441101, + "learning_rate": 3.575592058295017e-05, + "loss": 0.009363370016217232, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00941, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.09123861789703369, + "learning_rate": 3.549391545931585e-05, + "loss": 0.0007300444412976503, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00073, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.044113557785749435, + "learning_rate": 3.5232722063479914e-05, + "loss": 0.00047350398381240666, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.22140735387802124, + "learning_rate": 3.49723512647657e-05, + "loss": 0.0011342635843902826, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00113, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.12345008552074432, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0013970056315883994, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0014, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.02416493557393551, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.0003286268038209528, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00033, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.26387253403663635, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.001414710539393127, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00142, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.012582412920892239, + "learning_rate": 3.3939310220027456e-05, + "loss": 8.591653750045225e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00009, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.0038655376993119717, + "learning_rate": 3.3683214232914404e-05, + "loss": 5.268297536531463e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0408964529633522, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00027673851582221687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0886145755648613, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0005747403483837843, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.11402814835309982, + "learning_rate": 3.292029118616024e-05, + "loss": 0.000894754019100219, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.08461544662714005, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00066529237665236, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00067, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.2829553782939911, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0002811444574035704, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00028, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.03589304909110069, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00016819429583847523, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.16046081483364105, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0013975308975204825, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.2950722277164459, + "learning_rate": 3.166726850239794e-05, + "loss": 0.00403338810428977, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00404, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0004486274265218526, + "learning_rate": 3.141953535845912e-05, + "loss": 1.0177070180361625e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01730065606534481, + "learning_rate": 3.11727834939056e-05, + "loss": 7.366175123024732e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.4219278395175934, + "learning_rate": 3.092702317708967e-05, + "loss": 0.006937004160135984, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00696, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.008418221957981586, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.021447941544466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.3869379460811615, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.0042590186931192875, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00427, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.2765217125415802, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0020075999200344086, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00201, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.2115616798400879, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.0020617658738046885, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00206, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.011683800257742405, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.0001429952826583758, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00014, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0025901086628437042, + "learning_rate": 2.9473853553877484e-05, + "loss": 2.3585678718518466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.14587508141994476, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0011214152909815311, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00112, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.23, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.021550685167312622, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00018227534019388258, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00018, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.002367217792198062, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.783137279038783e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00003, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.02959338203072548, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.0001235282834386453, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.63, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.04836918041110039, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0002190673112636432, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.17935794591903687, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.001091235550120473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00109, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.17555589973926544, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0008262722985818982, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.015604093670845032, + "learning_rate": 2.7596140715257824e-05, + "loss": 7.283307058969513e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00007, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.001744552399031818, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.2661729670071509e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.03369157761335373, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.00013295823009684682, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00013, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.11042775958776474, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00030668824911117554, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00031, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.015241619199514389, + "learning_rate": 2.6684342539801933e-05, + "loss": 5.481348853209056e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.07663743942975998, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004268307238817215, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.3878970146179199, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.006505412515252829, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00653, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.011625024490058422, + "learning_rate": 2.6012832999747916e-05, + "loss": 7.441644265782088e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00007, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.8276505470275879, + "learning_rate": 2.579139666504821e-05, + "loss": 0.022178810089826584, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02243, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0010690620401874185, + "learning_rate": 2.557117581955798e-05, + "loss": 1.1797061233664863e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.06872054934501648, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0004525431140791625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00045, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.025393052026629448, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0001451169518986717, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.04, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.006796801462769508, + "learning_rate": 2.491789760603361e-05, + "loss": 9.397654503118247e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.35059845447540283, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.004584255628287792, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.33106186985969543, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.010121386498212814, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01017, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.003417555009946227, + "learning_rate": 2.427588563158384e-05, + "loss": 4.207541496725753e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.24088126420974731, + "learning_rate": 2.406442693028651e-05, + "loss": 0.004325853195041418, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00434, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.64, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.04677414149045944, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00039886703598313034, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0004, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.01845633052289486, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00019935239106416702, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.19680730998516083, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0020088059827685356, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.029754316434264183, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00013066174869891256, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00013, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.47083696722984314, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.006020379718393087, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00604, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.57, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.02095848135650158, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023951532784849405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.68, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.09784477949142456, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0008653341792523861, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00087, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.016663849353790283, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.0002501910785213113, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00025, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.06284014135599136, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0006274969782680273, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00063, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.06682170182466507, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0008653995464555919, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00087, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.048094261437654495, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0006572521524503827, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00066, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.288680762052536, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.0029174773953855038, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00292, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.009714324958622456, + "learning_rate": 2.1435742029681725e-05, + "loss": 8.201875607483089e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00939064659178257, + "learning_rate": 2.124308220868431e-05, + "loss": 0.00012157741002738476, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.035272955894470215, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0004544599214568734, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00045, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.22718806564807892, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0024650408886373043, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00247, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.013060114346444607, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.0001124740592786111, + "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": 31.69, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.052736442536115646, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0003598350449465215, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00036, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.00711337523534894, + "learning_rate": 2.0301012845054956e-05, + "loss": 8.437763608526438e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.00903877429664135, + "learning_rate": 2.011689980574966e-05, + "loss": 0.00010423596540931612, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0001, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.6410778164863586, + "learning_rate": 1.993423839463052e-05, + "loss": 0.004257278051227331, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00427, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.16, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.013556886464357376, + "learning_rate": 1.975303621298445e-05, + "loss": 0.0001645060838200152, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.10837272554636002, + "learning_rate": 1.957330080137385e-05, + "loss": 0.0001806107466109097, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.005454800557345152, + "learning_rate": 1.9395039639322864e-05, + "loss": 9.649700950831175e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.94, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.003915195353329182, + "learning_rate": 1.9218260145006073e-05, + "loss": 5.392331149778329e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.024293167516589165, + "learning_rate": 1.904296967493982e-05, + "loss": 0.00030749564757570624, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00031, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.1073823794722557, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0014971208292990923, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.0015, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.025999421253800392, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00022489691036753356, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.033106278628110886, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0003824468585662544, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00038, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.010407397523522377, + "learning_rate": 1.8356842992397304e-05, + "loss": 8.543908188585192e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00009, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.007676505949348211, + "learning_rate": 1.8189105812005714e-05, + "loss": 9.762673289515078e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0001, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.008882309310138226, + "learning_rate": 1.802290048317732e-05, + "loss": 0.00011588008783292025, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00012, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.78, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.01079952996224165, + "learning_rate": 1.785823392239424e-05, + "loss": 0.00015772700135130435, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00016, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.10464993119239807, + "learning_rate": 1.7695112982104225e-05, + "loss": 0.0007448438555002213, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00075, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.033524010330438614, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0003343333082739264, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00033, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.0037654207553714514, + "learning_rate": 1.7373535050913946e-05, + "loss": 4.5559077989310026e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00005, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.19, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.06157264485955238, + "learning_rate": 1.721509144218405e-05, + "loss": 0.0005529897753149271, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00055, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.009139563888311386, + "learning_rate": 1.705822021773101e-05, + "loss": 0.00013727511395700276, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.32006195187568665, + "learning_rate": 1.69029279056068e-05, + "loss": 0.005133518483489752, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00515, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.19554191827774048, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.0020202554296702147, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00202, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 39.09, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.010584031231701374, + "learning_rate": 1.659710580175893e-05, + "loss": 7.440667104674503e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.005364356562495232, + "learning_rate": 1.644658873654133e-05, + "loss": 6.896692502778023e-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": 37.33, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.0026337867602705956, + "learning_rate": 1.629767603613508e-05, + "loss": 5.40713153895922e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00005, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.85, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.08371307700872421, + "learning_rate": 1.615037389740547e-05, + "loss": 0.000647605920676142, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00065, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.2, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.12106083333492279, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0010971315205097198, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0011, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.0016024125507101417, + "learning_rate": 1.5860625757072092e-05, + "loss": 2.5408877263544127e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 433, + "tokens/total": 13106128, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.006628716830164194, + "learning_rate": 1.571819181307116e-05, + "loss": 9.017687989398837e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.023524144664406776, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0003021995653398335, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0003, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 37.86, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.1354794055223465, + "learning_rate": 1.543823381344311e-05, + "loss": 0.001777412835508585, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00178, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 38.06, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.0022158098872750998, + "learning_rate": 1.5300721408011114e-05, + "loss": 2.7318394131725654e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.005856889300048351, + "learning_rate": 1.5164861051607254e-05, + "loss": 7.72368221078068e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.007335511967539787, + "learning_rate": 1.5030658397935521e-05, + "loss": 7.384424679912627e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.00513842236250639, + "learning_rate": 1.4898119031716104e-05, + "loss": 8.217831782530993e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00008, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.0026071977335959673, + "learning_rate": 1.476724846845306e-05, + "loss": 3.909884981112555e-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": 40.13, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.0043723853304982185, + "learning_rate": 1.463805215420471e-05, + "loss": 4.3211588490521535e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.032961804419755936, + "learning_rate": 1.451053546535705e-05, + "loss": 0.00024515652330592275, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00025, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 35.82, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.12496478110551834, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0008977800025604665, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0009, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.64, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.016452984884381294, + "learning_rate": 1.4260562119706606e-05, + "loss": 0.00011743961658794433, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00012, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.58, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.009045973420143127, + "learning_rate": 1.413811586531508e-05, + "loss": 8.700705802766606e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00009, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.038517843931913376, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.0003440188302192837, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00034, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.019700370728969574, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.00021027779439464211, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00021, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.3, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..fe7868a2f43014278d9f8c1515865daea2f83d37 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c60dbf25f45e8c0c7399af9ef9cc0c6356455864304ee0d5363d0634cdfd6a5f +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..bde43b7d69c726c1793abcd11f1a0e64ba6b4799 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406d75229a30431561c43bc5b6c4184f215e56064b5a302bf8087430dec82177 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..9da896c9d8f84289d77b53501f8d2b2d3e8bc3ec --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce1c4e63a6c6f8be8c137d3add4eda184d59a043 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3da71c55273035b5e1d531117df684eced67a866 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..b28cebab2fbe968242be5b4d8b35002f886d4780 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.012836308218538761, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.00020321214105933905, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0002, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.01663857325911522, + "learning_rate": 4.90663667927174e-05, + "loss": 0.00027601540205068886, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.019089875742793083, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0003022018354386091, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.021187657490372658, + "learning_rate": 4.849136411748306e-05, + "loss": 0.00046957118320278823, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00047, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.21631699800491333, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.005231338553130627, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00525, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.74, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.06148816645145416, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0008393025491386652, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00084, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.96, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.05927408114075661, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0008574716048315167, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00086, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.019883394241333008, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.0003959762107115239, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.14877387881278992, + "learning_rate": 4.705880619288153e-05, + "loss": 0.002186706755310297, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00219, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.22, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 1.7873051166534424, + "learning_rate": 4.677323837567412e-05, + "loss": 0.004662086255848408, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00467, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.015957769006490707, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00035542104160413146, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00036, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.21242187917232513, + "learning_rate": 4.620314165804964e-05, + "loss": 0.002182526746764779, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00218, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.2956356704235077, + "learning_rate": 4.591863648168407e-05, + "loss": 0.007447251118719578, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00748, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.19968345761299133, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.002804415300488472, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00281, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.09324834495782852, + "learning_rate": 4.535077169046201e-05, + "loss": 0.001014861511066556, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00102, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.18735024333000183, + "learning_rate": 4.506743570677743e-05, + "loss": 0.0008113987278193235, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00081, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.20801964402198792, + "learning_rate": 4.478451305763618e-05, + "loss": 0.0025226804427802563, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00253, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.42, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.12481324374675751, + "learning_rate": 4.450201551660454e-05, + "loss": 0.001147121423855424, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00115, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.16472840309143066, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0030050487257540226, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00301, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.018171297386288643, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00021931837545707822, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.10386991500854492, + "learning_rate": 4.36571910095382e-05, + "loss": 0.0008659960003569722, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00087, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.18, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.020061908289790154, + "learning_rate": 4.337651127546448e-05, + "loss": 0.00014169455971568823, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00014, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.7, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.12189290672540665, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.0012950387317687273, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0013, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.00976351834833622, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.00013979707728140056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.2909281253814697, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.005748535506427288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00577, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.03204086795449257, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002670975518412888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.05219706520438194, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00025036506121978164, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00025, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.9549329876899719, + "learning_rate": 4.170299795992081e-05, + "loss": 0.01225995272397995, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01234, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.028625696897506714, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00029421402723528445, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00029, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.96, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.01120083313435316, + "learning_rate": 4.114946342220728e-05, + "loss": 0.00013539142673835158, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0018913366366177797, + "learning_rate": 4.087355496656321e-05, + "loss": 3.699818626046181e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.24486027657985687, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0035814084112644196, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00359, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.04, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.007224086206406355, + "learning_rate": 4.0323513089607876e-05, + "loss": 8.45451868372038e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00008, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.10228509455919266, + "learning_rate": 4.004940255778431e-05, + "loss": 0.0019895497243851423, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.08815925568342209, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0012562067713588476, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00126, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.04117776080965996, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00045941316056996584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00046, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.10311771184206009, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0007725170580670238, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00077, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0661429911851883, + "learning_rate": 3.895929566172861e-05, + "loss": 0.001271560788154602, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00127, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.48627135157585144, + "learning_rate": 3.868840945050728e-05, + "loss": 0.008270882070064545, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00831, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.12237530201673508, + "learning_rate": 3.841820203114995e-05, + "loss": 0.001470191404223442, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00147, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.1192391961812973, + "learning_rate": 3.814868464809027e-05, + "loss": 0.001252716057933867, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00125, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.5366963744163513, + "learning_rate": 3.787986851704667e-05, + "loss": 0.0043250457383692265, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00433, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.006924811284989119, + "learning_rate": 3.7611764824555654e-05, + "loss": 8.977689140010625e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.050131287425756454, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00035900447983294725, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00036, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.02984483353793621, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002764730015769601, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00028, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.004975741263478994, + "learning_rate": 3.68118397962661e-05, + "loss": 7.946819096105173e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.004153846297413111, + "learning_rate": 3.654669712344384e-05, + "loss": 5.989645796944387e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.005283186212182045, + "learning_rate": 3.628232236787763e-05, + "loss": 4.799446833203547e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.030474629253149033, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.00027237640460953116, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00027, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.59, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.5984165072441101, + "learning_rate": 3.575592058295017e-05, + "loss": 0.009363370016217232, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00941, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.09123861789703369, + "learning_rate": 3.549391545931585e-05, + "loss": 0.0007300444412976503, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00073, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.044113557785749435, + "learning_rate": 3.5232722063479914e-05, + "loss": 0.00047350398381240666, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.22140735387802124, + "learning_rate": 3.49723512647657e-05, + "loss": 0.0011342635843902826, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00113, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.12345008552074432, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0013970056315883994, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0014, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.02416493557393551, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.0003286268038209528, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00033, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.26387253403663635, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.001414710539393127, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00142, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.012582412920892239, + "learning_rate": 3.3939310220027456e-05, + "loss": 8.591653750045225e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00009, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.0038655376993119717, + "learning_rate": 3.3683214232914404e-05, + "loss": 5.268297536531463e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0408964529633522, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00027673851582221687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0886145755648613, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0005747403483837843, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.11402814835309982, + "learning_rate": 3.292029118616024e-05, + "loss": 0.000894754019100219, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.08461544662714005, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00066529237665236, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00067, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.2829553782939911, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0002811444574035704, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00028, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.03589304909110069, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00016819429583847523, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.16046081483364105, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0013975308975204825, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.2950722277164459, + "learning_rate": 3.166726850239794e-05, + "loss": 0.00403338810428977, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00404, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0004486274265218526, + "learning_rate": 3.141953535845912e-05, + "loss": 1.0177070180361625e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01730065606534481, + "learning_rate": 3.11727834939056e-05, + "loss": 7.366175123024732e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.4219278395175934, + "learning_rate": 3.092702317708967e-05, + "loss": 0.006937004160135984, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00696, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.008418221957981586, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.021447941544466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.3869379460811615, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.0042590186931192875, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00427, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.2765217125415802, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0020075999200344086, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00201, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.2115616798400879, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.0020617658738046885, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00206, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.011683800257742405, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.0001429952826583758, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00014, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0025901086628437042, + "learning_rate": 2.9473853553877484e-05, + "loss": 2.3585678718518466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.14587508141994476, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0011214152909815311, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00112, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.23, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.021550685167312622, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00018227534019388258, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00018, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.002367217792198062, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.783137279038783e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00003, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.02959338203072548, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.0001235282834386453, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.63, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.04836918041110039, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0002190673112636432, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.17935794591903687, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.001091235550120473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00109, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.17555589973926544, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0008262722985818982, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.015604093670845032, + "learning_rate": 2.7596140715257824e-05, + "loss": 7.283307058969513e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00007, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.001744552399031818, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.2661729670071509e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.03369157761335373, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.00013295823009684682, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00013, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.11042775958776474, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00030668824911117554, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00031, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.015241619199514389, + "learning_rate": 2.6684342539801933e-05, + "loss": 5.481348853209056e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.07663743942975998, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004268307238817215, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.3878970146179199, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.006505412515252829, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00653, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.011625024490058422, + "learning_rate": 2.6012832999747916e-05, + "loss": 7.441644265782088e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00007, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.8276505470275879, + "learning_rate": 2.579139666504821e-05, + "loss": 0.022178810089826584, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02243, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0010690620401874185, + "learning_rate": 2.557117581955798e-05, + "loss": 1.1797061233664863e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.06872054934501648, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0004525431140791625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00045, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.025393052026629448, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0001451169518986717, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.04, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.006796801462769508, + "learning_rate": 2.491789760603361e-05, + "loss": 9.397654503118247e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.35059845447540283, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.004584255628287792, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.33106186985969543, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.010121386498212814, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01017, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.003417555009946227, + "learning_rate": 2.427588563158384e-05, + "loss": 4.207541496725753e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.24088126420974731, + "learning_rate": 2.406442693028651e-05, + "loss": 0.004325853195041418, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00434, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.64, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.04677414149045944, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00039886703598313034, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0004, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.01845633052289486, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00019935239106416702, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.19680730998516083, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0020088059827685356, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.029754316434264183, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00013066174869891256, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00013, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.47083696722984314, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.006020379718393087, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00604, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.57, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.02095848135650158, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023951532784849405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.68, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.09784477949142456, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0008653341792523861, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00087, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.016663849353790283, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.0002501910785213113, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00025, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.06284014135599136, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0006274969782680273, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00063, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.06682170182466507, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0008653995464555919, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00087, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.048094261437654495, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0006572521524503827, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00066, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.288680762052536, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.0029174773953855038, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00292, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.009714324958622456, + "learning_rate": 2.1435742029681725e-05, + "loss": 8.201875607483089e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00939064659178257, + "learning_rate": 2.124308220868431e-05, + "loss": 0.00012157741002738476, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.035272955894470215, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0004544599214568734, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00045, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.22718806564807892, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0024650408886373043, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00247, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.013060114346444607, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.0001124740592786111, + "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": 31.69, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.052736442536115646, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0003598350449465215, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00036, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.00711337523534894, + "learning_rate": 2.0301012845054956e-05, + "loss": 8.437763608526438e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.00903877429664135, + "learning_rate": 2.011689980574966e-05, + "loss": 0.00010423596540931612, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0001, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.6410778164863586, + "learning_rate": 1.993423839463052e-05, + "loss": 0.004257278051227331, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00427, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.16, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.013556886464357376, + "learning_rate": 1.975303621298445e-05, + "loss": 0.0001645060838200152, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.10837272554636002, + "learning_rate": 1.957330080137385e-05, + "loss": 0.0001806107466109097, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.005454800557345152, + "learning_rate": 1.9395039639322864e-05, + "loss": 9.649700950831175e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.94, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.003915195353329182, + "learning_rate": 1.9218260145006073e-05, + "loss": 5.392331149778329e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.024293167516589165, + "learning_rate": 1.904296967493982e-05, + "loss": 0.00030749564757570624, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00031, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.1073823794722557, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0014971208292990923, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.0015, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.025999421253800392, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00022489691036753356, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.033106278628110886, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0003824468585662544, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00038, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.010407397523522377, + "learning_rate": 1.8356842992397304e-05, + "loss": 8.543908188585192e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00009, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.007676505949348211, + "learning_rate": 1.8189105812005714e-05, + "loss": 9.762673289515078e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0001, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.008882309310138226, + "learning_rate": 1.802290048317732e-05, + "loss": 0.00011588008783292025, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00012, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.78, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.01079952996224165, + "learning_rate": 1.785823392239424e-05, + "loss": 0.00015772700135130435, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00016, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.10464993119239807, + "learning_rate": 1.7695112982104225e-05, + "loss": 0.0007448438555002213, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00075, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.033524010330438614, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0003343333082739264, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00033, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.0037654207553714514, + "learning_rate": 1.7373535050913946e-05, + "loss": 4.5559077989310026e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00005, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.19, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.06157264485955238, + "learning_rate": 1.721509144218405e-05, + "loss": 0.0005529897753149271, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00055, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.009139563888311386, + "learning_rate": 1.705822021773101e-05, + "loss": 0.00013727511395700276, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.32006195187568665, + "learning_rate": 1.69029279056068e-05, + "loss": 0.005133518483489752, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00515, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.19554191827774048, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.0020202554296702147, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00202, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 39.09, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.010584031231701374, + "learning_rate": 1.659710580175893e-05, + "loss": 7.440667104674503e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.005364356562495232, + "learning_rate": 1.644658873654133e-05, + "loss": 6.896692502778023e-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": 37.33, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.0026337867602705956, + "learning_rate": 1.629767603613508e-05, + "loss": 5.40713153895922e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00005, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.85, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.08371307700872421, + "learning_rate": 1.615037389740547e-05, + "loss": 0.000647605920676142, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00065, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.2, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.12106083333492279, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0010971315205097198, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0011, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.0016024125507101417, + "learning_rate": 1.5860625757072092e-05, + "loss": 2.5408877263544127e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 433, + "tokens/total": 13106128, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.006628716830164194, + "learning_rate": 1.571819181307116e-05, + "loss": 9.017687989398837e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.023524144664406776, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0003021995653398335, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0003, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 37.86, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.1354794055223465, + "learning_rate": 1.543823381344311e-05, + "loss": 0.001777412835508585, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00178, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 38.06, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.0022158098872750998, + "learning_rate": 1.5300721408011114e-05, + "loss": 2.7318394131725654e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.005856889300048351, + "learning_rate": 1.5164861051607254e-05, + "loss": 7.72368221078068e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.007335511967539787, + "learning_rate": 1.5030658397935521e-05, + "loss": 7.384424679912627e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.00513842236250639, + "learning_rate": 1.4898119031716104e-05, + "loss": 8.217831782530993e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00008, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.0026071977335959673, + "learning_rate": 1.476724846845306e-05, + "loss": 3.909884981112555e-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": 40.13, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.0043723853304982185, + "learning_rate": 1.463805215420471e-05, + "loss": 4.3211588490521535e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.032961804419755936, + "learning_rate": 1.451053546535705e-05, + "loss": 0.00024515652330592275, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00025, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 35.82, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.12496478110551834, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0008977800025604665, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0009, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.64, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.016452984884381294, + "learning_rate": 1.4260562119706606e-05, + "loss": 0.00011743961658794433, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00012, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.58, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.009045973420143127, + "learning_rate": 1.413811586531508e-05, + "loss": 8.700705802766606e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00009, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.038517843931913376, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.0003440188302192837, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00034, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.019700370728969574, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.00021027779439464211, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00021, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 204821 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.0006826155004091561, + "learning_rate": 1.3780999708818058e-05, + "loss": 1.4103720786806662e-05, + "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": 33.92, + "tokens/trainable": 205271 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.008357573300600052, + "learning_rate": 1.3665385037857758e-05, + "loss": 9.161404886981472e-05, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 450, + "tokens/total": 13620960, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 205723 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.01647551730275154, + "learning_rate": 1.3551490468947126e-05, + "loss": 0.00017527706222608685, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00018, + "step": 451, + "tokens/total": 13651344, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 206169 + }, + { + "epoch": 1.765625, + "grad_norm": 0.03474664315581322, + "learning_rate": 1.3439320741704075e-05, + "loss": 0.0004909664858132601, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00049, + "step": 452, + "tokens/total": 13681792, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 206592 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.005533512681722641, + "learning_rate": 1.3328880523968808e-05, + "loss": 8.479113603243604e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00008, + "step": 453, + "tokens/total": 13712256, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 207055 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.028869982808828354, + "learning_rate": 1.3220174411609587e-05, + "loss": 7.264575106091797e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00007, + "step": 454, + "tokens/total": 13742128, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 207494 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.0076578096486628056, + "learning_rate": 1.3113206928331471e-05, + "loss": 2.4288401618832722e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 455, + "tokens/total": 13772160, + "tokens/train_per_sec_per_gpu": 33.22, + "tokens/trainable": 207956 + }, + { + "epoch": 1.78125, + "grad_norm": 0.0006251306622289121, + "learning_rate": 1.300798252548806e-05, + "loss": 1.4033696061233059e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00001, + "step": 456, + "tokens/total": 13802448, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 208408 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.0268402099609375, + "learning_rate": 1.2904505581896265e-05, + "loss": 0.00014731910778209567, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00015, + "step": 457, + "tokens/total": 13832880, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 208889 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.060711171478033066, + "learning_rate": 1.2802780403654082e-05, + "loss": 0.00020443923131097108, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0002, + "step": 458, + "tokens/total": 13862880, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 209336 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.12638092041015625, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.0009473193786107004, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00095, + "step": 459, + "tokens/total": 13893440, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 209797 + }, + { + "epoch": 1.796875, + "grad_norm": 0.0015253991587087512, + "learning_rate": 1.2604602202943861e-05, + "loss": 2.2379259462468326e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00002, + "step": 460, + "tokens/total": 13923824, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 210240 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.013747433200478554, + "learning_rate": 1.2508157427479686e-05, + "loss": 9.997214510804042e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0001, + "step": 461, + "tokens/total": 13954208, + "tokens/train_per_sec_per_gpu": 35.22, + "tokens/trainable": 210702 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.08502306789159775, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.0007195118232630193, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00072, + "step": 462, + "tokens/total": 13984784, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 211140 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.016438405960798264, + "learning_rate": 1.2320576593470082e-05, + "loss": 9.172871796181425e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00009, + "step": 463, + "tokens/total": 14014896, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 211575 + }, + { + "epoch": 1.8125, + "grad_norm": 0.09976232051849365, + "learning_rate": 1.2229448340928828e-05, + "loss": 0.000811650650575757, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00081, + "step": 464, + "tokens/total": 14045312, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 212083 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.00435422919690609, + "learning_rate": 1.2140099945624458e-05, + "loss": 2.5664587155915797e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 465, + "tokens/total": 14075840, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 212558 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.0026768327224999666, + "learning_rate": 1.205253512570841e-05, + "loss": 3.4629334550118074e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 466, + "tokens/total": 14106016, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 213014 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0026948358863592148, + "learning_rate": 1.1966757525110255e-05, + "loss": 4.344901753938757e-05, + "memory/device_reserved (GiB)": 37.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00004, + "step": 467, + "tokens/total": 14136512, + "tokens/train_per_sec_per_gpu": 34.3, + "tokens/trainable": 213465 + }, + { + "epoch": 1.828125, + "grad_norm": 0.006161310710012913, + "learning_rate": 1.1882770713386095e-05, + "loss": 4.4334883568808436e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 468, + "tokens/total": 14166832, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 213940 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.010646386072039604, + "learning_rate": 1.180057818556998e-05, + "loss": 2.2965485186432488e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00002, + "step": 469, + "tokens/total": 14197360, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 214415 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.0007588419830426574, + "learning_rate": 1.1720183362028494e-05, + "loss": 1.3413812666840386e-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": 34.61, + "tokens/trainable": 214880 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.017162201926112175, + "learning_rate": 1.1641589588318387e-05, + "loss": 0.0001371123653370887, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 471, + "tokens/total": 14257888, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 215334 + }, + { + "epoch": 1.84375, + "grad_norm": 0.0008721952908672392, + "learning_rate": 1.1564800135047418e-05, + "loss": 1.1102063581347466e-05, + "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": 35.27, + "tokens/trainable": 215822 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.029533211141824722, + "learning_rate": 1.148981819773816e-05, + "loss": 0.0002879542880691588, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 473, + "tokens/total": 14318448, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 216239 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.0012399809202179313, + "learning_rate": 1.1416646896695086e-05, + "loss": 1.8869191990233958e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 474, + "tokens/total": 14348912, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 216691 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.22101251780986786, + "learning_rate": 1.1345289276874717e-05, + "loss": 0.0024789171293377876, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00248, + "step": 475, + "tokens/total": 14379440, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 217162 + }, + { + "epoch": 1.859375, + "grad_norm": 0.004043657332658768, + "learning_rate": 1.1275748307758873e-05, + "loss": 4.4153628550702706e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 476, + "tokens/total": 14409840, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 217638 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.022147774696350098, + "learning_rate": 1.1208026883231147e-05, + "loss": 0.00015884214371908456, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00016, + "step": 477, + "tokens/total": 14440256, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 218053 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.0010556846391409636, + "learning_rate": 1.1142127821456433e-05, + "loss": 1.8117327272193506e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 478, + "tokens/total": 14470704, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 218503 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.019303888082504272, + "learning_rate": 1.1078053864763674e-05, + "loss": 0.0001503409876022488, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00015, + "step": 479, + "tokens/total": 14501072, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 218964 + }, + { + "epoch": 1.875, + "grad_norm": 0.008807498030364513, + "learning_rate": 1.1015807679531756e-05, + "loss": 6.010168726788834e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00006, + "step": 480, + "tokens/total": 14531200, + "tokens/train_per_sec_per_gpu": 35.51, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..7a8c5aaccac9e3fd9891693d672fab4ab00b8391 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:573d247ce118e2fb451bfb4f5b9042c0a5fb28ade8fc70e42ab3f968627dcc23 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..a6808004e207bd8163746b966322b12399b5b6b1 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe0e31a2ce5f388d7493cfd427bdb094831beb2135f230f975d487f8c82fe0a6 +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..0626d83bfee73afaca3ed469625b1cf5142c7a61 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..22325b3b3d08e0697784fe21b8c321485a51f7b5 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..115a5bf49aff4472c7ae208d22e2b298d69dbe1c --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..5cf07c20db90aa4b549f57d6ee5d786a64f5f132 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 43865 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.1577538549900055, + "learning_rate": 9.53619478457953e-05, + "loss": 0.006561641581356525, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00658, + "step": 97, + "tokens/total": 2932272, + "tokens/train_per_sec_per_gpu": 32.26, + "tokens/trainable": 44328 + }, + { + "epoch": 0.3828125, + "grad_norm": 0.2749426066875458, + "learning_rate": 9.523275153154695e-05, + "loss": 0.007920589298009872, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00795, + "step": 98, + "tokens/total": 2962032, + "tokens/train_per_sec_per_gpu": 31.3, + "tokens/trainable": 44778 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.47395026683807373, + "learning_rate": 9.51018809682839e-05, + "loss": 0.03451812267303467, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.03512, + "step": 99, + "tokens/total": 2992256, + "tokens/train_per_sec_per_gpu": 37.58, + "tokens/trainable": 45225 + }, + { + "epoch": 0.390625, + "grad_norm": 0.5378239750862122, + "learning_rate": 9.49693416020645e-05, + "loss": 0.017406146973371506, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 100, + "tokens/total": 3022608, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 45678 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.45281967520713806, + "learning_rate": 9.483513894839276e-05, + "loss": 0.004519028589129448, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00453, + "step": 101, + "tokens/total": 3052992, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 46125 + }, + { + "epoch": 0.3984375, + "grad_norm": 1.7491872310638428, + "learning_rate": 9.469927859198888e-05, + "loss": 0.00577917555347085, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0058, + "step": 102, + "tokens/total": 3083392, + "tokens/train_per_sec_per_gpu": 40.05, + "tokens/trainable": 46612 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.3060601055622101, + "learning_rate": 9.456176618655689e-05, + "loss": 0.008232416585087776, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00827, + "step": 103, + "tokens/total": 3113744, + "tokens/train_per_sec_per_gpu": 32.96, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.7810171246528625, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03279104456305504, + "memory/device_reserved (GiB)": 35.8, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.03333, + "step": 104, + "tokens/total": 3144272, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 47536 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.4873018264770508, + "learning_rate": 9.428180818692884e-05, + "loss": 0.008343804627656937, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00838, + "step": 105, + "tokens/total": 3174512, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 48037 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.26075872778892517, + "learning_rate": 9.413937424292791e-05, + "loss": 0.00894663855433464, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00899, + "step": 106, + "tokens/total": 3204896, + "tokens/train_per_sec_per_gpu": 32.94, + "tokens/trainable": 48491 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.9285218715667725, + "learning_rate": 9.399531154980424e-05, + "loss": 0.0386667475104332, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.03942, + "step": 107, + "tokens/total": 3235360, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 48940 + }, + { + "epoch": 0.421875, + "grad_norm": 0.37687361240386963, + "learning_rate": 9.384962610259455e-05, + "loss": 0.014155298471450806, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01426, + "step": 108, + "tokens/total": 3265552, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 49389 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.12366248667240143, + "learning_rate": 9.370232396386494e-05, + "loss": 0.0033611906692385674, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00337, + "step": 109, + "tokens/total": 3293856, + "tokens/train_per_sec_per_gpu": 36.26, + "tokens/trainable": 49838 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.3310869038105011, + "learning_rate": 9.355341126345868e-05, + "loss": 0.01677614450454712, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01692, + "step": 110, + "tokens/total": 3323984, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 50310 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.16138073801994324, + "learning_rate": 9.340289419824107e-05, + "loss": 0.005432053469121456, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00545, + "step": 111, + "tokens/total": 3354320, + "tokens/train_per_sec_per_gpu": 33.94, + "tokens/trainable": 50755 + }, + { + "epoch": 0.4375, + "grad_norm": 0.2602401673793793, + "learning_rate": 9.325077903184159e-05, + "loss": 0.009463014081120491, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00951, + "step": 112, + "tokens/total": 3384880, + "tokens/train_per_sec_per_gpu": 31.85, + "tokens/trainable": 51213 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.11247953772544861, + "learning_rate": 9.30970720943932e-05, + "loss": 0.0028001072350889444, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0028, + "step": 113, + "tokens/total": 3415104, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 51660 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.2779226303100586, + "learning_rate": 9.2941779782269e-05, + "loss": 0.01312212087213993, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01321, + "step": 114, + "tokens/total": 3445504, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 52133 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.22592218220233917, + "learning_rate": 9.278490855781596e-05, + "loss": 0.005544666200876236, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 115, + "tokens/total": 3475744, + "tokens/train_per_sec_per_gpu": 37.85, + "tokens/trainable": 52580 + }, + { + "epoch": 0.453125, + "grad_norm": 0.272401362657547, + "learning_rate": 9.262646494908604e-05, + "loss": 0.005348580423742533, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00536, + "step": 116, + "tokens/total": 3506016, + "tokens/train_per_sec_per_gpu": 35.2, + "tokens/trainable": 53033 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.9178272485733032, + "learning_rate": 9.246645554956457e-05, + "loss": 0.022306114435195923, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02256, + "step": 117, + "tokens/total": 3536256, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 53517 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.10259439796209335, + "learning_rate": 9.230488701789578e-05, + "loss": 0.0009993483545258641, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.001, + "step": 118, + "tokens/total": 3566480, + "tokens/train_per_sec_per_gpu": 34.96, + "tokens/trainable": 53967 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.15369175374507904, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0015791363548487425, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00158, + "step": 119, + "tokens/total": 3596624, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 54430 + }, + { + "epoch": 0.46875, + "grad_norm": 0.4007995128631592, + "learning_rate": 9.197709951682268e-05, + "loss": 0.008637797087430954, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00868, + "step": 120, + "tokens/total": 3627168, + "tokens/train_per_sec_per_gpu": 35.57, + "tokens/trainable": 54896 + }, + { + "epoch": 0.47265625, + "grad_norm": 2.9844160079956055, + "learning_rate": 9.181089418799428e-05, + "loss": 0.01763911172747612, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0178, + "step": 121, + "tokens/total": 3657632, + "tokens/train_per_sec_per_gpu": 38.12, + "tokens/trainable": 55379 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.4593406021595001, + "learning_rate": 9.164315700760271e-05, + "loss": 0.007832051254808903, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00786, + "step": 122, + "tokens/total": 3687824, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 55803 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.22288177907466888, + "learning_rate": 9.147389495587671e-05, + "loss": 0.0024647461250424385, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00247, + "step": 123, + "tokens/total": 3718320, + "tokens/train_per_sec_per_gpu": 36.8, + "tokens/trainable": 56249 + }, + { + "epoch": 0.484375, + "grad_norm": 3.4173779487609863, + "learning_rate": 9.130311507650116e-05, + "loss": 0.010984445922076702, + "memory/device_reserved (GiB)": 37.62, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01104, + "step": 124, + "tokens/total": 3748672, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 56713 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.46252161264419556, + "learning_rate": 9.113082447632394e-05, + "loss": 0.0051902830600738525, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0052, + "step": 125, + "tokens/total": 3778976, + "tokens/train_per_sec_per_gpu": 39.56, + "tokens/trainable": 57196 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.3998625874519348, + "learning_rate": 9.09570303250602e-05, + "loss": 0.004260494839400053, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00427, + "step": 126, + "tokens/total": 3809296, + "tokens/train_per_sec_per_gpu": 36.93, + "tokens/trainable": 57680 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.23200765252113342, + "learning_rate": 9.078173985499394e-05, + "loss": 0.003244318999350071, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00325, + "step": 127, + "tokens/total": 3839328, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 58110 + }, + { + "epoch": 0.5, + "grad_norm": 0.26066046953201294, + "learning_rate": 9.060496036067713e-05, + "loss": 0.011201048269867897, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01126, + "step": 128, + "tokens/total": 3869824, + "tokens/train_per_sec_per_gpu": 32.02, + "tokens/trainable": 58534 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.7281484007835388, + "learning_rate": 9.042669919862615e-05, + "loss": 0.011813497170805931, + "memory/device_reserved (GiB)": 37.76, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01188, + "step": 129, + "tokens/total": 3900080, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 58998 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.5487067699432373, + "learning_rate": 9.024696378701557e-05, + "loss": 0.008090369403362274, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00812, + "step": 130, + "tokens/total": 3930560, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 59448 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.5761887431144714, + "learning_rate": 9.006576160536948e-05, + "loss": 0.008387972600758076, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00842, + "step": 131, + "tokens/total": 3960496, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 59906 + }, + { + "epoch": 0.515625, + "grad_norm": 0.34462451934814453, + "learning_rate": 8.988310019425035e-05, + "loss": 0.005629569757729769, + "memory/device_reserved (GiB)": 35.93, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00565, + "step": 132, + "tokens/total": 3990928, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 60350 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.7096816897392273, + "learning_rate": 8.969898715494506e-05, + "loss": 0.010645382106304169, + "memory/device_reserved (GiB)": 37.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0107, + "step": 133, + "tokens/total": 4021264, + "tokens/train_per_sec_per_gpu": 36.48, + "tokens/trainable": 60831 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.9463919997215271, + "learning_rate": 8.951343014914869e-05, + "loss": 0.018044205382466316, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01821, + "step": 134, + "tokens/total": 4051728, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 61305 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.20147432386875153, + "learning_rate": 8.932643689864568e-05, + "loss": 0.00538910087198019, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.0054, + "step": 135, + "tokens/total": 4081920, + "tokens/train_per_sec_per_gpu": 38.01, + "tokens/trainable": 61792 + }, + { + "epoch": 0.53125, + "grad_norm": 0.1833702027797699, + "learning_rate": 8.913801518498845e-05, + "loss": 0.00374586577527225, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00375, + "step": 136, + "tokens/total": 4112304, + "tokens/train_per_sec_per_gpu": 31.67, + "tokens/trainable": 62253 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.7401099801063538, + "learning_rate": 8.894817284917364e-05, + "loss": 0.01421094499528408, + "memory/device_reserved (GiB)": 38.55, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01431, + "step": 137, + "tokens/total": 4142512, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 62707 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.2769958972930908, + "learning_rate": 8.875691779131569e-05, + "loss": 0.0012122916523367167, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00121, + "step": 138, + "tokens/total": 4172992, + "tokens/train_per_sec_per_gpu": 29.6, + "tokens/trainable": 63145 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.2656077444553375, + "learning_rate": 8.856425797031829e-05, + "loss": 0.005337266717106104, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00535, + "step": 139, + "tokens/total": 4203136, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 63587 + }, + { + "epoch": 0.546875, + "grad_norm": 0.15931734442710876, + "learning_rate": 8.837020140354295e-05, + "loss": 0.003361564828082919, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00337, + "step": 140, + "tokens/total": 4233456, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 64052 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.2797258496284485, + "learning_rate": 8.817475616647554e-05, + "loss": 0.006030731834471226, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00605, + "step": 141, + "tokens/total": 4263728, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 64526 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.18650899827480316, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0034353849478065968, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00344, + "step": 142, + "tokens/total": 4294432, + "tokens/train_per_sec_per_gpu": 34.97, + "tokens/trainable": 64983 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.18854950368404388, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004699597600847483, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00471, + "step": 143, + "tokens/total": 4324960, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 65492 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1250438392162323, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0013500247150659561, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00135, + "step": 144, + "tokens/total": 4355424, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 65925 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.2933589518070221, + "learning_rate": 8.737925204046629e-05, + "loss": 0.008707849308848381, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00875, + "step": 145, + "tokens/total": 4385712, + "tokens/train_per_sec_per_gpu": 31.73, + "tokens/trainable": 66383 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.30792802572250366, + "learning_rate": 8.717698659491851e-05, + "loss": 0.005249719135463238, + "memory/device_reserved (GiB)": 38.56, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00526, + "step": 146, + "tokens/total": 4416016, + "tokens/train_per_sec_per_gpu": 31.65, + "tokens/trainable": 66840 + }, + { + "epoch": 0.57421875, + "grad_norm": 1.029360055923462, + "learning_rate": 8.697338213361735e-05, + "loss": 0.014694154262542725, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0148, + "step": 147, + "tokens/total": 4446368, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 67297 + }, + { + "epoch": 0.578125, + "grad_norm": 0.2955895662307739, + "learning_rate": 8.676844712937552e-05, + "loss": 0.004935478325933218, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00495, + "step": 148, + "tokens/total": 4476880, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 67769 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.556270956993103, + "learning_rate": 8.656219011037509e-05, + "loss": 0.023344792425632477, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02362, + "step": 149, + "tokens/total": 4507232, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 68257 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.43625330924987793, + "learning_rate": 8.63546196598125e-05, + "loss": 0.009602912701666355, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00965, + "step": 150, + "tokens/total": 4537376, + "tokens/train_per_sec_per_gpu": 35.86, + "tokens/trainable": 68706 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.2960459291934967, + "learning_rate": 8.614574441554145e-05, + "loss": 0.009531187824904919, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00958, + "step": 151, + "tokens/total": 4567584, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 69131 + }, + { + "epoch": 0.59375, + "grad_norm": 0.22226394712924957, + "learning_rate": 8.593557306971349e-05, + "loss": 0.013231074437499046, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01332, + "step": 152, + "tokens/total": 4597792, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 69586 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.13233518600463867, + "learning_rate": 8.572411436841618e-05, + "loss": 0.0031906149815768003, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0032, + "step": 153, + "tokens/total": 4627936, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 70061 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.17915832996368408, + "learning_rate": 8.551137711130922e-05, + "loss": 0.006568522658199072, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00659, + "step": 154, + "tokens/total": 4658352, + "tokens/train_per_sec_per_gpu": 28.23, + "tokens/trainable": 70467 + }, + { + "epoch": 0.60546875, + "grad_norm": 0.3126574456691742, + "learning_rate": 8.529737015125824e-05, + "loss": 0.014315147884190083, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01442, + "step": 155, + "tokens/total": 4688896, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 70933 + }, + { + "epoch": 0.609375, + "grad_norm": 0.19957849383354187, + "learning_rate": 8.508210239396639e-05, + "loss": 0.0077377320267260075, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00777, + "step": 156, + "tokens/total": 4719168, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 71382 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.07674852013587952, + "learning_rate": 8.486558279760375e-05, + "loss": 0.0022812150418758392, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00228, + "step": 157, + "tokens/total": 4749136, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 71808 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.29424574971199036, + "learning_rate": 8.464782037243449e-05, + "loss": 0.009841009974479675, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00989, + "step": 158, + "tokens/total": 4779472, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 72249 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.38520634174346924, + "learning_rate": 8.442882418044202e-05, + "loss": 0.017565667629241943, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01772, + "step": 159, + "tokens/total": 4809632, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 72726 + }, + { + "epoch": 0.625, + "grad_norm": 0.1961541771888733, + "learning_rate": 8.420860333495179e-05, + "loss": 0.00602555088698864, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00604, + "step": 160, + "tokens/total": 4840112, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 73148 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.3786018490791321, + "learning_rate": 8.398716700025208e-05, + "loss": 0.0033199097961187363, + "memory/device_reserved (GiB)": 38.65, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00333, + "step": 161, + "tokens/total": 4870656, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 73600 + }, + { + "epoch": 0.6328125, + "grad_norm": 1.5304594039916992, + "learning_rate": 8.376452439121266e-05, + "loss": 0.013961974531412125, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01406, + "step": 162, + "tokens/total": 4900608, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 74068 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.3360394835472107, + "learning_rate": 8.354068477290124e-05, + "loss": 0.010342447087168694, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.0104, + "step": 163, + "tokens/total": 4930752, + "tokens/train_per_sec_per_gpu": 36.1, + "tokens/trainable": 74527 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1671362817287445, + "learning_rate": 8.331565746019807e-05, + "loss": 0.006741576828062534, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00676, + "step": 164, + "tokens/total": 4961008, + "tokens/train_per_sec_per_gpu": 32.16, + "tokens/trainable": 74992 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.338470995426178, + "learning_rate": 8.308945181740812e-05, + "loss": 0.009520573541522026, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00957, + "step": 165, + "tokens/total": 4991200, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 75442 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.06734886765480042, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0011861161328852177, + "memory/device_reserved (GiB)": 36.0, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00119, + "step": 166, + "tokens/total": 5021600, + "tokens/train_per_sec_per_gpu": 31.54, + "tokens/trainable": 75887 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.5632230043411255, + "learning_rate": 8.263354324357182e-05, + "loss": 0.02382841520011425, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02411, + "step": 167, + "tokens/total": 5052032, + "tokens/train_per_sec_per_gpu": 34.45, + "tokens/trainable": 76331 + }, + { + "epoch": 0.65625, + "grad_norm": 0.14913584291934967, + "learning_rate": 8.240385928474219e-05, + "loss": 0.002766846213489771, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00277, + "step": 168, + "tokens/total": 5080256, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 76745 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.1904655545949936, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0046575069427490234, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00467, + "step": 169, + "tokens/total": 5110784, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 77201 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.3416374921798706, + "learning_rate": 8.194107981329746e-05, + "loss": 0.007014934439212084, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00704, + "step": 170, + "tokens/total": 5141200, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 77655 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.06584338843822479, + "learning_rate": 8.170800355882518e-05, + "loss": 0.0010388410883024335, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00104, + "step": 171, + "tokens/total": 5171760, + "tokens/train_per_sec_per_gpu": 39.64, + "tokens/trainable": 78122 + }, + { + "epoch": 0.671875, + "grad_norm": 1.4229398965835571, + "learning_rate": 8.147381587530713e-05, + "loss": 0.015414456836879253, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01553, + "step": 172, + "tokens/total": 5202272, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 78576 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3806024193763733, + "learning_rate": 8.123852650824877e-05, + "loss": 0.012036355212330818, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01211, + "step": 173, + "tokens/total": 5232800, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 79059 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.5877760052680969, + "learning_rate": 8.100214524900103e-05, + "loss": 0.01602815091609955, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.01616, + "step": 174, + "tokens/total": 5262672, + "tokens/train_per_sec_per_gpu": 29.99, + "tokens/trainable": 79498 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.33037039637565613, + "learning_rate": 8.076468193435301e-05, + "loss": 0.010002637282013893, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01005, + "step": 175, + "tokens/total": 5293072, + "tokens/train_per_sec_per_gpu": 30.73, + "tokens/trainable": 79922 + }, + { + "epoch": 0.6875, + "grad_norm": 0.4363207221031189, + "learning_rate": 8.052614644612253e-05, + "loss": 0.016971714794635773, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.01712, + "step": 176, + "tokens/total": 5321520, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 80383 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.40556958317756653, + "learning_rate": 8.028654871074489e-05, + "loss": 0.00855817086994648, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00859, + "step": 177, + "tokens/total": 5351904, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 80824 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.16822558641433716, + "learning_rate": 8.004589869885986e-05, + "loss": 0.006781051866710186, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 178, + "tokens/total": 5382432, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 81243 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.28983059525489807, + "learning_rate": 7.980420642489674e-05, + "loss": 0.005999822169542313, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00602, + "step": 179, + "tokens/total": 5412960, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 81716 + }, + { + "epoch": 0.703125, + "grad_norm": 0.18040503561496735, + "learning_rate": 7.95614819466576e-05, + "loss": 0.004937571007758379, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00495, + "step": 180, + "tokens/total": 5443232, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 82181 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.3842947483062744, + "learning_rate": 7.931773536489872e-05, + "loss": 0.014881981536746025, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.01499, + "step": 181, + "tokens/total": 5471440, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 82626 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.1560991108417511, + "learning_rate": 7.907297682291035e-05, + "loss": 0.004242315422743559, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00425, + "step": 182, + "tokens/total": 5501744, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 83089 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.1657864898443222, + "learning_rate": 7.882721650609442e-05, + "loss": 0.005256622098386288, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00527, + "step": 183, + "tokens/total": 5532272, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 83590 + }, + { + "epoch": 0.71875, + "grad_norm": 0.6097539067268372, + "learning_rate": 7.85804646415409e-05, + "loss": 0.016960084438323975, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0171, + "step": 184, + "tokens/total": 5562784, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 84046 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.09047499299049377, + "learning_rate": 7.833273149760207e-05, + "loss": 0.002440867479890585, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00244, + "step": 185, + "tokens/total": 5592800, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 84517 + }, + { + "epoch": 0.7265625, + "grad_norm": 3.610114574432373, + "learning_rate": 7.808402738346527e-05, + "loss": 0.027849921956658363, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02824, + "step": 186, + "tokens/total": 5623088, + "tokens/train_per_sec_per_gpu": 34.34, + "tokens/trainable": 84974 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.09014768898487091, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0016205484280362725, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00162, + "step": 187, + "tokens/total": 5653344, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 85460 + }, + { + "epoch": 0.734375, + "grad_norm": 0.11981519311666489, + "learning_rate": 7.758374768294647e-05, + "loss": 0.004210016690194607, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 188, + "tokens/total": 5683600, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 85939 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.21320489048957825, + "learning_rate": 7.733219291524489e-05, + "loss": 0.0067794108763337135, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0068, + "step": 189, + "tokens/total": 5711952, + "tokens/train_per_sec_per_gpu": 38.83, + "tokens/trainable": 86377 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.07177300751209259, + "learning_rate": 7.707970881383977e-05, + "loss": 0.0022305003367364407, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00223, + "step": 190, + "tokens/total": 5742336, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 86834 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.8523200154304504, + "learning_rate": 7.682630588562518e-05, + "loss": 0.00609443848952651, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00611, + "step": 191, + "tokens/total": 5772560, + "tokens/train_per_sec_per_gpu": 32.42, + "tokens/trainable": 87265 + }, + { + "epoch": 0.75, + "grad_norm": 0.24030189216136932, + "learning_rate": 7.657199467573129e-05, + "loss": 0.004537718836218119, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00455, + "step": 192, + "tokens/total": 5803136, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 87747 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.17005977034568787, + "learning_rate": 7.631678576708561e-05, + "loss": 0.003625418059527874, + "memory/device_reserved (GiB)": 38.58, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00363, + "step": 193, + "tokens/total": 5833440, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 88239 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.22860349714756012, + "learning_rate": 7.606068977997255e-05, + "loss": 0.0007644169381819665, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00076, + "step": 194, + "tokens/total": 5863552, + "tokens/train_per_sec_per_gpu": 36.27, + "tokens/trainable": 88718 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.5368777513504028, + "learning_rate": 7.580371737159148e-05, + "loss": 0.015955395996570587, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01608, + "step": 195, + "tokens/total": 5893680, + "tokens/train_per_sec_per_gpu": 31.42, + "tokens/trainable": 89129 + }, + { + "epoch": 0.765625, + "grad_norm": 0.04851720482110977, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0004464306985028088, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00045, + "step": 196, + "tokens/total": 5923744, + "tokens/train_per_sec_per_gpu": 32.58, + "tokens/trainable": 89567 + }, + { + "epoch": 0.76953125, + "grad_norm": 2.685831308364868, + "learning_rate": 7.528718610173511e-05, + "loss": 0.03162113204598427, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03213, + "step": 197, + "tokens/total": 5954128, + "tokens/train_per_sec_per_gpu": 30.71, + "tokens/trainable": 89988 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.11396117508411407, + "learning_rate": 7.502764873523431e-05, + "loss": 0.0007245016167871654, + "memory/device_reserved (GiB)": 35.56, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00072, + "step": 198, + "tokens/total": 5984352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 90434 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.4947444498538971, + "learning_rate": 7.476727793652011e-05, + "loss": 0.01100912131369114, + "memory/device_reserved (GiB)": 35.6, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01107, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 34.51, + "tokens/trainable": 90913 + }, + { + "epoch": 0.78125, + "grad_norm": 0.6466076374053955, + "learning_rate": 7.450608454068415e-05, + "loss": 0.01665530353784561, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.01679, + "step": 200, + "tokens/total": 6045056, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 91355 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.33924102783203125, + "learning_rate": 7.424407941704987e-05, + "loss": 0.008527595549821854, + "memory/device_reserved (GiB)": 35.66, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00856, + "step": 201, + "tokens/total": 6075504, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 91856 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.11541225016117096, + "learning_rate": 7.398127346871986e-05, + "loss": 0.001006019883789122, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00101, + "step": 202, + "tokens/total": 6105904, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 92311 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.013573888689279556, + "learning_rate": 7.371767763212238e-05, + "loss": 0.00013932021101936698, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00014, + "step": 203, + "tokens/total": 6136272, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 92764 + }, + { + "epoch": 0.796875, + "grad_norm": 0.3160369098186493, + "learning_rate": 7.345330287655617e-05, + "loss": 0.013633212074637413, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.01373, + "step": 204, + "tokens/total": 6166336, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 93227 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.2614762783050537, + "learning_rate": 7.31881602037339e-05, + "loss": 0.006595024839043617, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00662, + "step": 205, + "tokens/total": 6196720, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 93675 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.2725606858730316, + "learning_rate": 7.29222606473245e-05, + "loss": 0.005595955532044172, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00561, + "step": 206, + "tokens/total": 6227424, + "tokens/train_per_sec_per_gpu": 32.68, + "tokens/trainable": 94120 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.05208409205079079, + "learning_rate": 7.265561527249383e-05, + "loss": 0.0006884766044095159, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00069, + "step": 207, + "tokens/total": 6257616, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 94557 + }, + { + "epoch": 0.8125, + "grad_norm": 0.6857002973556519, + "learning_rate": 7.238823517544436e-05, + "loss": 0.010781912133097649, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01084, + "step": 208, + "tokens/total": 6288000, + "tokens/train_per_sec_per_gpu": 40.97, + "tokens/trainable": 95035 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.18070466816425323, + "learning_rate": 7.212013148295333e-05, + "loss": 0.002757919952273369, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00276, + "step": 209, + "tokens/total": 6318336, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 95517 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.16365700960159302, + "learning_rate": 7.185131535190975e-05, + "loss": 0.005094663705676794, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00511, + "step": 210, + "tokens/total": 6348656, + "tokens/train_per_sec_per_gpu": 36.74, + "tokens/trainable": 96026 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2167346179485321, + "learning_rate": 7.158179796885005e-05, + "loss": 0.007257625460624695, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00728, + "step": 211, + "tokens/total": 6379040, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 96477 + }, + { + "epoch": 0.828125, + "grad_norm": 0.08415517956018448, + "learning_rate": 7.131159054949273e-05, + "loss": 0.0018573766574263573, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00186, + "step": 212, + "tokens/total": 6409312, + "tokens/train_per_sec_per_gpu": 33.39, + "tokens/trainable": 96951 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.4260450601577759, + "learning_rate": 7.104070433827139e-05, + "loss": 0.021458052098751068, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02169, + "step": 213, + "tokens/total": 6439520, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 97350 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.13258714973926544, + "learning_rate": 7.076915060786705e-05, + "loss": 0.0031774642411619425, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00318, + "step": 214, + "tokens/total": 6469888, + "tokens/train_per_sec_per_gpu": 30.04, + "tokens/trainable": 97792 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.4242945909500122, + "learning_rate": 7.049694065873882e-05, + "loss": 0.017319604754447937, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01747, + "step": 215, + "tokens/total": 6500128, + "tokens/train_per_sec_per_gpu": 36.65, + "tokens/trainable": 98257 + }, + { + "epoch": 0.84375, + "grad_norm": 0.3156379461288452, + "learning_rate": 7.022408581865382e-05, + "loss": 0.015216855332255363, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.01533, + "step": 216, + "tokens/total": 6530832, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 98731 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.05867718160152435, + "learning_rate": 6.99505974422157e-05, + "loss": 0.00046426919288933277, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00046, + "step": 217, + "tokens/total": 6561248, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 99212 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.0972680002450943, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0036130475345999002, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00362, + "step": 218, + "tokens/total": 6591648, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 99654 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.10315816849470139, + "learning_rate": 6.940176563004123e-05, + "loss": 0.0020681945607066154, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00207, + "step": 219, + "tokens/total": 6622368, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 100086 + }, + { + "epoch": 0.859375, + "grad_norm": 0.09276516735553741, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0020987873431295156, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0021, + "step": 220, + "tokens/total": 6652848, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 100524 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.3432714343070984, + "learning_rate": 6.885053657779273e-05, + "loss": 0.006101437844336033, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00612, + "step": 221, + "tokens/total": 6683392, + "tokens/train_per_sec_per_gpu": 33.63, + "tokens/trainable": 100996 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.32484498620033264, + "learning_rate": 6.857405174478604e-05, + "loss": 0.005491882096976042, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00551, + "step": 222, + "tokens/total": 6713712, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 101449 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.408407986164093, + "learning_rate": 6.82970020400792e-05, + "loss": 0.014964321628212929, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01508, + "step": 223, + "tokens/total": 6744176, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 101905 + }, + { + "epoch": 0.875, + "grad_norm": 0.07530955970287323, + "learning_rate": 6.801939899284132e-05, + "loss": 0.002201129449531436, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0022, + "step": 224, + "tokens/total": 6774416, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 102411 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.07113490253686905, + "learning_rate": 6.774125415526827e-05, + "loss": 0.001676997751928866, + "memory/device_reserved (GiB)": 37.19, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00168, + "step": 225, + "tokens/total": 6804592, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 102882 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.02237725257873535, + "learning_rate": 6.746257910210214e-05, + "loss": 0.0004955717595294118, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0005, + "step": 226, + "tokens/total": 6834976, + "tokens/train_per_sec_per_gpu": 34.85, + "tokens/trainable": 103332 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.2688956558704376, + "learning_rate": 6.718338543014937e-05, + "loss": 0.01203202549368143, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0121, + "step": 227, + "tokens/total": 6865408, + "tokens/train_per_sec_per_gpu": 35.35, + "tokens/trainable": 103790 + }, + { + "epoch": 0.890625, + "grad_norm": 0.11738378554582596, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0008421347592957318, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00084, + "step": 228, + "tokens/total": 6895664, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 104246 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.7972850203514099, + "learning_rate": 6.662348872453553e-05, + "loss": 0.006434948183596134, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00646, + "step": 229, + "tokens/total": 6926016, + "tokens/train_per_sec_per_gpu": 37.92, + "tokens/trainable": 104707 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.037981245666742325, + "learning_rate": 6.63428089904618e-05, + "loss": 0.0003998051688540727, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 230, + "tokens/total": 6956384, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 105167 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.06838542222976685, + "learning_rate": 6.60616572358065e-05, + "loss": 0.0011449911398813128, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00115, + "step": 231, + "tokens/total": 6986768, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 105576 + }, + { + "epoch": 0.90625, + "grad_norm": 0.0395381897687912, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0007394620915874839, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00074, + "step": 232, + "tokens/total": 7017232, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 106043 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.40814921259880066, + "learning_rate": 6.549798448339548e-05, + "loss": 0.005036160349845886, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00505, + "step": 233, + "tokens/total": 7047568, + "tokens/train_per_sec_per_gpu": 35.52, + "tokens/trainable": 106506 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.03264355659484863, + "learning_rate": 6.521548694236384e-05, + "loss": 0.0003136758750770241, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00031, + "step": 234, + "tokens/total": 7077760, + "tokens/train_per_sec_per_gpu": 34.49, + "tokens/trainable": 106951 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.06884738057851791, + "learning_rate": 6.493256429322259e-05, + "loss": 0.000555745791643858, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00056, + "step": 235, + "tokens/total": 7107760, + "tokens/train_per_sec_per_gpu": 31.02, + "tokens/trainable": 107382 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1841747760772705, + "learning_rate": 6.464922830953799e-05, + "loss": 0.003819565288722515, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00383, + "step": 236, + "tokens/total": 7138144, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 107814 + }, + { + "epoch": 0.92578125, + "grad_norm": 1.5293514728546143, + "learning_rate": 6.436549078207688e-05, + "loss": 0.02590152993798256, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02624, + "step": 237, + "tokens/total": 7168352, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 108274 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.06575962156057358, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0009253169409930706, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00093, + "step": 238, + "tokens/total": 7198624, + "tokens/train_per_sec_per_gpu": 29.35, + "tokens/trainable": 108714 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.048577819019556046, + "learning_rate": 6.379685834195036e-05, + "loss": 0.0002521112619433552, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00025, + "step": 239, + "tokens/total": 7229216, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 109220 + }, + { + "epoch": 0.9375, + "grad_norm": 0.28133440017700195, + "learning_rate": 6.351198709240186e-05, + "loss": 0.003380296751856804, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00339, + "step": 240, + "tokens/total": 7259648, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 109672 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.30106109380722046, + "learning_rate": 6.32267616243259e-05, + "loss": 0.0028490356635302305, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00285, + "step": 241, + "tokens/total": 7289824, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 110135 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.4230552911758423, + "learning_rate": 6.294119380711849e-05, + "loss": 0.004133709240704775, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00414, + "step": 242, + "tokens/total": 7320288, + "tokens/train_per_sec_per_gpu": 30.81, + "tokens/trainable": 110563 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.2737712860107422, + "learning_rate": 6.265529552442209e-05, + "loss": 0.0038032641168683767, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00381, + "step": 243, + "tokens/total": 7350736, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 111049 + }, + { + "epoch": 0.953125, + "grad_norm": 0.16202203929424286, + "learning_rate": 6.236907867363127e-05, + "loss": 0.0011738601606339216, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 244, + "tokens/total": 7381280, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 111495 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.4818413555622101, + "learning_rate": 6.208255516539749e-05, + "loss": 0.011270345188677311, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01133, + "step": 245, + "tokens/total": 7411632, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 111968 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.09653587639331818, + "learning_rate": 6.179573692313344e-05, + "loss": 0.0007583714323118329, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00076, + "step": 246, + "tokens/total": 7441904, + "tokens/train_per_sec_per_gpu": 36.76, + "tokens/trainable": 112416 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.07513346523046494, + "learning_rate": 6.150863588251694e-05, + "loss": 0.0008964500157162547, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0009, + "step": 247, + "tokens/total": 7472368, + "tokens/train_per_sec_per_gpu": 34.22, + "tokens/trainable": 112882 + }, + { + "epoch": 0.96875, + "grad_norm": 0.069392628967762, + "learning_rate": 6.122126399099419e-05, + "loss": 0.001440772321075201, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00144, + "step": 248, + "tokens/total": 7502656, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 113390 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.48798877000808716, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.010719450190663338, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01078, + "step": 249, + "tokens/total": 7532800, + "tokens/train_per_sec_per_gpu": 38.0, + "tokens/trainable": 113904 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.2219225913286209, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0012265645200386643, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00123, + "step": 250, + "tokens/total": 7563264, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 114351 + }, + { + "epoch": 0.98046875, + "grad_norm": 2.7423243522644043, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.004063127096742392, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00407, + "step": 251, + "tokens/total": 7593840, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 114842 + }, + { + "epoch": 0.984375, + "grad_norm": 0.3986411690711975, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.004339811857789755, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00435, + "step": 252, + "tokens/total": 7624416, + "tokens/train_per_sec_per_gpu": 29.1, + "tokens/trainable": 115263 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.23857781291007996, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.003166551236063242, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00317, + "step": 253, + "tokens/total": 7654688, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 115703 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.13943128287792206, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.002166214631870389, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00217, + "step": 254, + "tokens/total": 7685088, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 116157 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.5217611193656921, + "learning_rate": 5.920308275189541e-05, + "loss": 0.011248281225562096, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01131, + "step": 255, + "tokens/total": 7715552, + "tokens/train_per_sec_per_gpu": 32.28, + "tokens/trainable": 116567 + }, + { + "epoch": 1.0, + "grad_norm": 0.12602734565734863, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.002226680051535368, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00223, + "step": 256, + "tokens/total": 7745872, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 117030 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.06412868201732635, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0004693535447586328, + "memory/device_reserved (GiB)": 38.23, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00047, + "step": 257, + "tokens/total": 7776208, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 117517 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.21263156831264496, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0017973874928429723, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0018, + "step": 258, + "tokens/total": 7806480, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 118015 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.16070301830768585, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026787673123180866, + "memory/device_reserved (GiB)": 35.62, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00268, + "step": 259, + "tokens/total": 7836800, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 118520 + }, + { + "epoch": 1.015625, + "grad_norm": 0.09660778194665909, + "learning_rate": 5.77560376810977e-05, + "loss": 0.0018036369001492858, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00181, + "step": 260, + "tokens/total": 7867040, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 118992 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.4718175232410431, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.007852890528738499, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00788, + "step": 261, + "tokens/total": 7897408, + "tokens/train_per_sec_per_gpu": 34.06, + "tokens/trainable": 119438 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.2906242311000824, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0019686920568346977, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00197, + "step": 262, + "tokens/total": 7927648, + "tokens/train_per_sec_per_gpu": 32.53, + "tokens/trainable": 119881 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.01843550242483616, + "learning_rate": 5.688633799118971e-05, + "loss": 0.0003263683756813407, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00033, + "step": 263, + "tokens/total": 7957968, + "tokens/train_per_sec_per_gpu": 28.15, + "tokens/trainable": 120285 + }, + { + "epoch": 1.03125, + "grad_norm": 0.16429829597473145, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0017057846998795867, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00171, + "step": 264, + "tokens/total": 7987888, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 120755 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.0684906542301178, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.0006418414413928986, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00064, + "step": 265, + "tokens/total": 8018048, + "tokens/train_per_sec_per_gpu": 35.94, + "tokens/trainable": 121194 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.4636438488960266, + "learning_rate": 5.601593183686955e-05, + "loss": 0.007831420749425888, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00786, + "step": 266, + "tokens/total": 8048448, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 121670 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.2955789566040039, + "learning_rate": 5.572569579717961e-05, + "loss": 0.009791678749024868, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00984, + "step": 267, + "tokens/total": 8078848, + "tokens/train_per_sec_per_gpu": 34.02, + "tokens/trainable": 122142 + }, + { + "epoch": 1.046875, + "grad_norm": 0.31020334362983704, + "learning_rate": 5.543542955832538e-05, + "loss": 0.004724925849586725, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00474, + "step": 268, + "tokens/total": 8109360, + "tokens/train_per_sec_per_gpu": 32.73, + "tokens/trainable": 122585 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.08568736165761948, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0010347942588850856, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00104, + "step": 269, + "tokens/total": 8139888, + "tokens/train_per_sec_per_gpu": 39.26, + "tokens/trainable": 123091 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.09669848531484604, + "learning_rate": 5.485485480053015e-05, + "loss": 0.001207374152727425, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00121, + "step": 270, + "tokens/total": 8170480, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 123505 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.19564303755760193, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0023664506152272224, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.00237, + "step": 271, + "tokens/total": 8198848, + "tokens/train_per_sec_per_gpu": 35.83, + "tokens/trainable": 123953 + }, + { + "epoch": 1.0625, + "grad_norm": 0.37416645884513855, + "learning_rate": 5.42743042028204e-05, + "loss": 0.014112848788499832, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01421, + "step": 272, + "tokens/total": 8229200, + "tokens/train_per_sec_per_gpu": 28.96, + "tokens/trainable": 124400 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.49067050218582153, + "learning_rate": 5.3984068163130464e-05, + "loss": 0.021221842616796494, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02145, + "step": 273, + "tokens/total": 8259536, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 124870 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.241201713681221, + "learning_rate": 5.369387440051111e-05, + "loss": 0.006090693175792694, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 274, + "tokens/total": 8289904, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 125333 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.019017836079001427, + "learning_rate": 5.340373499110935e-05, + "loss": 0.0002696830197237432, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00027, + "step": 275, + "tokens/total": 8320176, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 125834 + }, + { + "epoch": 1.078125, + "grad_norm": 0.011502111330628395, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.00019440105825196952, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00019, + "step": 276, + "tokens/total": 8350832, + "tokens/train_per_sec_per_gpu": 38.45, + "tokens/trainable": 126316 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.26989492774009705, + "learning_rate": 5.282366752473479e-05, + "loss": 0.006379636470228434, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0064, + "step": 277, + "tokens/total": 8380976, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 126798 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.8612332940101624, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.01699192076921463, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01714, + "step": 278, + "tokens/total": 8411072, + "tokens/train_per_sec_per_gpu": 29.66, + "tokens/trainable": 127223 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.32299262285232544, + "learning_rate": 5.224396231890232e-05, + "loss": 0.00695072254166007, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.00697, + "step": 279, + "tokens/total": 8440736, + "tokens/train_per_sec_per_gpu": 36.28, + "tokens/trainable": 127672 + }, + { + "epoch": 1.09375, + "grad_norm": 0.12054527550935745, + "learning_rate": 5.195427572104522e-05, + "loss": 0.0043365005403757095, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00435, + "step": 280, + "tokens/total": 8470944, + "tokens/train_per_sec_per_gpu": 32.49, + "tokens/trainable": 128116 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.10208329558372498, + "learning_rate": 5.166471586820751e-05, + "loss": 0.001871119486168027, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00187, + "step": 281, + "tokens/total": 8501104, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 128589 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.20265866816043854, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.006930817384272814, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00695, + "step": 282, + "tokens/total": 8531696, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 129036 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.30612000823020935, + "learning_rate": 5.1086024590884144e-05, + "loss": 0.009662694297730923, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00971, + "step": 283, + "tokens/total": 8561936, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 129485 + }, + { + "epoch": 1.109375, + "grad_norm": 0.06664866954088211, + "learning_rate": 5.079691724810461e-05, + "loss": 0.0016388238873332739, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00164, + "step": 284, + "tokens/total": 8592192, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 129920 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.38304245471954346, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002592553850263357, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0026, + "step": 285, + "tokens/total": 8622432, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 130417 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.06439055502414703, + "learning_rate": 5.021923930849237e-05, + "loss": 0.0009901091689243913, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00099, + "step": 286, + "tokens/total": 8652768, + "tokens/train_per_sec_per_gpu": 36.33, + "tokens/trainable": 130903 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.14527085423469543, + "learning_rate": 4.99306927511967e-05, + "loss": 0.002050987211987376, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00205, + "step": 287, + "tokens/total": 8683296, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 131343 + }, + { + "epoch": 1.125, + "grad_norm": 0.14959585666656494, + "learning_rate": 4.964235714846775e-05, + "loss": 0.0029775684233754873, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00298, + "step": 288, + "tokens/total": 8713504, + "tokens/train_per_sec_per_gpu": 32.55, + "tokens/trainable": 131763 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.012836308218538761, + "learning_rate": 4.9354244499126866e-05, + "loss": 0.00020321214105933905, + "memory/device_reserved (GiB)": 35.98, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0002, + "step": 289, + "tokens/total": 8743888, + "tokens/train_per_sec_per_gpu": 34.28, + "tokens/trainable": 132229 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.01663857325911522, + "learning_rate": 4.90663667927174e-05, + "loss": 0.00027601540205068886, + "memory/device_reserved (GiB)": 35.54, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 290, + "tokens/total": 8774336, + "tokens/train_per_sec_per_gpu": 35.72, + "tokens/trainable": 132678 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.019089875742793083, + "learning_rate": 4.877873600900581e-05, + "loss": 0.0003022018354386091, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0003, + "step": 291, + "tokens/total": 8804816, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 133136 + }, + { + "epoch": 1.140625, + "grad_norm": 0.021187657490372658, + "learning_rate": 4.849136411748306e-05, + "loss": 0.00046957118320278823, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00047, + "step": 292, + "tokens/total": 8835024, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 133608 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.21631699800491333, + "learning_rate": 4.8204263076866574e-05, + "loss": 0.005231338553130627, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00525, + "step": 293, + "tokens/total": 8865408, + "tokens/train_per_sec_per_gpu": 40.74, + "tokens/trainable": 134108 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.06148816645145416, + "learning_rate": 4.791744483460251e-05, + "loss": 0.0008393025491386652, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00084, + "step": 294, + "tokens/total": 8895824, + "tokens/train_per_sec_per_gpu": 30.96, + "tokens/trainable": 134541 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.05927408114075661, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0008574716048315167, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00086, + "step": 295, + "tokens/total": 8926336, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 134966 + }, + { + "epoch": 1.15625, + "grad_norm": 0.019883394241333008, + "learning_rate": 4.7344704475577916e-05, + "loss": 0.0003959762107115239, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0004, + "step": 296, + "tokens/total": 8956656, + "tokens/train_per_sec_per_gpu": 33.77, + "tokens/trainable": 135402 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.14877387881278992, + "learning_rate": 4.705880619288153e-05, + "loss": 0.002186706755310297, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00219, + "step": 297, + "tokens/total": 8986752, + "tokens/train_per_sec_per_gpu": 30.22, + "tokens/trainable": 135804 + }, + { + "epoch": 1.1640625, + "grad_norm": 1.7873051166534424, + "learning_rate": 4.677323837567412e-05, + "loss": 0.004662086255848408, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00467, + "step": 298, + "tokens/total": 9017024, + "tokens/train_per_sec_per_gpu": 31.34, + "tokens/trainable": 136231 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.015957769006490707, + "learning_rate": 4.6488012907598146e-05, + "loss": 0.00035542104160413146, + "memory/device_reserved (GiB)": 35.55, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00036, + "step": 299, + "tokens/total": 9047424, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 136705 + }, + { + "epoch": 1.171875, + "grad_norm": 0.21242187917232513, + "learning_rate": 4.620314165804964e-05, + "loss": 0.002182526746764779, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00218, + "step": 300, + "tokens/total": 9077648, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 137178 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.2956356704235077, + "learning_rate": 4.591863648168407e-05, + "loss": 0.007447251118719578, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00748, + "step": 301, + "tokens/total": 9108032, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 137643 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.19968345761299133, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.002804415300488472, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00281, + "step": 302, + "tokens/total": 9138240, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 138078 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.09324834495782852, + "learning_rate": 4.535077169046201e-05, + "loss": 0.001014861511066556, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00102, + "step": 303, + "tokens/total": 9168352, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 138515 + }, + { + "epoch": 1.1875, + "grad_norm": 0.18735024333000183, + "learning_rate": 4.506743570677743e-05, + "loss": 0.0008113987278193235, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00081, + "step": 304, + "tokens/total": 9198864, + "tokens/train_per_sec_per_gpu": 35.45, + "tokens/trainable": 138948 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.20801964402198792, + "learning_rate": 4.478451305763618e-05, + "loss": 0.0025226804427802563, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00253, + "step": 305, + "tokens/total": 9229104, + "tokens/train_per_sec_per_gpu": 38.42, + "tokens/trainable": 139408 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.12481324374675751, + "learning_rate": 4.450201551660454e-05, + "loss": 0.001147121423855424, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00115, + "step": 306, + "tokens/total": 9257344, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 139840 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.16472840309143066, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.0030050487257540226, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00301, + "step": 307, + "tokens/total": 9287520, + "tokens/train_per_sec_per_gpu": 33.2, + "tokens/trainable": 140281 + }, + { + "epoch": 1.203125, + "grad_norm": 0.018171297386288643, + "learning_rate": 4.393834276419352e-05, + "loss": 0.00021931837545707822, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 308, + "tokens/total": 9317888, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 140776 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.10386991500854492, + "learning_rate": 4.36571910095382e-05, + "loss": 0.0008659960003569722, + "memory/device_reserved (GiB)": 35.72, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00087, + "step": 309, + "tokens/total": 9348256, + "tokens/train_per_sec_per_gpu": 37.18, + "tokens/trainable": 141228 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.020061908289790154, + "learning_rate": 4.337651127546448e-05, + "loss": 0.00014169455971568823, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00014, + "step": 310, + "tokens/total": 9378560, + "tokens/train_per_sec_per_gpu": 37.7, + "tokens/trainable": 141679 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.12189290672540665, + "learning_rate": 4.3096315242201736e-05, + "loss": 0.0012950387317687273, + "memory/device_reserved (GiB)": 35.74, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.0013, + "step": 311, + "tokens/total": 9408848, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 142122 + }, + { + "epoch": 1.21875, + "grad_norm": 0.00976351834833622, + "learning_rate": 4.2816614569850635e-05, + "loss": 0.00013979707728140056, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00014, + "step": 312, + "tokens/total": 9439344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 142561 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.2909281253814697, + "learning_rate": 4.2537420897897864e-05, + "loss": 0.005748535506427288, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00577, + "step": 313, + "tokens/total": 9469792, + "tokens/train_per_sec_per_gpu": 35.97, + "tokens/trainable": 143046 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.03204086795449257, + "learning_rate": 4.225874584473174e-05, + "loss": 0.0002670975518412888, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00027, + "step": 314, + "tokens/total": 9500128, + "tokens/train_per_sec_per_gpu": 33.26, + "tokens/trainable": 143493 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.05219706520438194, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00025036506121978164, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00025, + "step": 315, + "tokens/total": 9530480, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 143956 + }, + { + "epoch": 1.234375, + "grad_norm": 0.9549329876899719, + "learning_rate": 4.170299795992081e-05, + "loss": 0.01225995272397995, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01234, + "step": 316, + "tokens/total": 9560912, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 144411 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.028625696897506714, + "learning_rate": 4.142594825521398e-05, + "loss": 0.00029421402723528445, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00029, + "step": 317, + "tokens/total": 9591344, + "tokens/train_per_sec_per_gpu": 38.96, + "tokens/trainable": 144892 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.01120083313435316, + "learning_rate": 4.114946342220728e-05, + "loss": 0.00013539142673835158, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 318, + "tokens/total": 9621392, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 145339 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.0018913366366177797, + "learning_rate": 4.087355496656321e-05, + "loss": 3.699818626046181e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 319, + "tokens/total": 9651728, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 145811 + }, + { + "epoch": 1.25, + "grad_norm": 0.24486027657985687, + "learning_rate": 4.05982343699588e-05, + "loss": 0.0035814084112644196, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00359, + "step": 320, + "tokens/total": 9682224, + "tokens/train_per_sec_per_gpu": 39.04, + "tokens/trainable": 146314 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.007224086206406355, + "learning_rate": 4.0323513089607876e-05, + "loss": 8.45451868372038e-05, + "memory/device_reserved (GiB)": 37.95, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00008, + "step": 321, + "tokens/total": 9712656, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 146781 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.10228509455919266, + "learning_rate": 4.004940255778431e-05, + "loss": 0.0019895497243851423, + "memory/device_reserved (GiB)": 34.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00199, + "step": 322, + "tokens/total": 9743088, + "tokens/train_per_sec_per_gpu": 35.81, + "tokens/trainable": 147225 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.08815925568342209, + "learning_rate": 3.977591418134619e-05, + "loss": 0.0012562067713588476, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00126, + "step": 323, + "tokens/total": 9773808, + "tokens/train_per_sec_per_gpu": 35.1, + "tokens/trainable": 147673 + }, + { + "epoch": 1.265625, + "grad_norm": 0.04117776080965996, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00045941316056996584, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00046, + "step": 324, + "tokens/total": 9804016, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 148103 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.10311771184206009, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0007725170580670238, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00077, + "step": 325, + "tokens/total": 9834592, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 148535 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0661429911851883, + "learning_rate": 3.895929566172861e-05, + "loss": 0.001271560788154602, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00127, + "step": 326, + "tokens/total": 9864848, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 148999 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.48627135157585144, + "learning_rate": 3.868840945050728e-05, + "loss": 0.008270882070064545, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00831, + "step": 327, + "tokens/total": 9895168, + "tokens/train_per_sec_per_gpu": 31.92, + "tokens/trainable": 149431 + }, + { + "epoch": 1.28125, + "grad_norm": 0.12237530201673508, + "learning_rate": 3.841820203114995e-05, + "loss": 0.001470191404223442, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00147, + "step": 328, + "tokens/total": 9925696, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 149886 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.1192391961812973, + "learning_rate": 3.814868464809027e-05, + "loss": 0.001252716057933867, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00125, + "step": 329, + "tokens/total": 9955648, + "tokens/train_per_sec_per_gpu": 31.31, + "tokens/trainable": 150288 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.5366963744163513, + "learning_rate": 3.787986851704667e-05, + "loss": 0.0043250457383692265, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00433, + "step": 330, + "tokens/total": 9985856, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 150733 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.006924811284989119, + "learning_rate": 3.7611764824555654e-05, + "loss": 8.977689140010625e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00009, + "step": 331, + "tokens/total": 10016208, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 151207 + }, + { + "epoch": 1.296875, + "grad_norm": 0.050131287425756454, + "learning_rate": 3.734438472750619e-05, + "loss": 0.00035900447983294725, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00036, + "step": 332, + "tokens/total": 10046512, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 151694 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.02984483353793621, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0002764730015769601, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00028, + "step": 333, + "tokens/total": 10076944, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 152188 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.004975741263478994, + "learning_rate": 3.68118397962661e-05, + "loss": 7.946819096105173e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00008, + "step": 334, + "tokens/total": 10107296, + "tokens/train_per_sec_per_gpu": 30.9, + "tokens/trainable": 152596 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.004153846297413111, + "learning_rate": 3.654669712344384e-05, + "loss": 5.989645796944387e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 335, + "tokens/total": 10137568, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 153052 + }, + { + "epoch": 1.3125, + "grad_norm": 0.005283186212182045, + "learning_rate": 3.628232236787763e-05, + "loss": 4.799446833203547e-05, + "memory/device_reserved (GiB)": 37.17, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 336, + "tokens/total": 10167952, + "tokens/train_per_sec_per_gpu": 30.43, + "tokens/trainable": 153491 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.030474629253149033, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.00027237640460953116, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00027, + "step": 337, + "tokens/total": 10198512, + "tokens/train_per_sec_per_gpu": 40.59, + "tokens/trainable": 153983 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.5984165072441101, + "learning_rate": 3.575592058295017e-05, + "loss": 0.009363370016217232, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00941, + "step": 338, + "tokens/total": 10228752, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 154464 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.09123861789703369, + "learning_rate": 3.549391545931585e-05, + "loss": 0.0007300444412976503, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00073, + "step": 339, + "tokens/total": 10259104, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 154924 + }, + { + "epoch": 1.328125, + "grad_norm": 0.044113557785749435, + "learning_rate": 3.5232722063479914e-05, + "loss": 0.00047350398381240666, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00047, + "step": 340, + "tokens/total": 10289520, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 155373 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.22140735387802124, + "learning_rate": 3.49723512647657e-05, + "loss": 0.0011342635843902826, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00113, + "step": 341, + "tokens/total": 10320016, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 155873 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.12345008552074432, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0013970056315883994, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0014, + "step": 342, + "tokens/total": 10350368, + "tokens/train_per_sec_per_gpu": 30.7, + "tokens/trainable": 156278 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.02416493557393551, + "learning_rate": 3.4454120764386764e-05, + "loss": 0.0003286268038209528, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00033, + "step": 343, + "tokens/total": 10380624, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 156733 + }, + { + "epoch": 1.34375, + "grad_norm": 0.26387253403663635, + "learning_rate": 3.4196282628408526e-05, + "loss": 0.001414710539393127, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00142, + "step": 344, + "tokens/total": 10411024, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 157203 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.012582412920892239, + "learning_rate": 3.3939310220027456e-05, + "loss": 8.591653750045225e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00009, + "step": 345, + "tokens/total": 10441248, + "tokens/train_per_sec_per_gpu": 38.09, + "tokens/trainable": 157707 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.0038655376993119717, + "learning_rate": 3.3683214232914404e-05, + "loss": 5.268297536531463e-05, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00005, + "step": 346, + "tokens/total": 10471712, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 158180 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0408964529633522, + "learning_rate": 3.342800532426873e-05, + "loss": 0.00027673851582221687, + "memory/device_reserved (GiB)": 38.13, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00028, + "step": 347, + "tokens/total": 10502288, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 158646 + }, + { + "epoch": 1.359375, + "grad_norm": 0.0886145755648613, + "learning_rate": 3.317369411437484e-05, + "loss": 0.0005747403483837843, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00057, + "step": 348, + "tokens/total": 10532304, + "tokens/train_per_sec_per_gpu": 35.88, + "tokens/trainable": 159115 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.11402814835309982, + "learning_rate": 3.292029118616024e-05, + "loss": 0.000894754019100219, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0009, + "step": 349, + "tokens/total": 10562608, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 159538 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.08461544662714005, + "learning_rate": 3.266780708475511e-05, + "loss": 0.00066529237665236, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00067, + "step": 350, + "tokens/total": 10592992, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 160016 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.2829553782939911, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0002811444574035704, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00028, + "step": 351, + "tokens/total": 10623328, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 160475 + }, + { + "epoch": 1.375, + "grad_norm": 0.03589304909110069, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00016819429583847523, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00017, + "step": 352, + "tokens/total": 10653632, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 160952 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.16046081483364105, + "learning_rate": 3.191597261653475e-05, + "loss": 0.0013975308975204825, + "memory/device_reserved (GiB)": 38.4, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0014, + "step": 353, + "tokens/total": 10684176, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 161420 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.2950722277164459, + "learning_rate": 3.166726850239794e-05, + "loss": 0.00403338810428977, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00404, + "step": 354, + "tokens/total": 10714352, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 161868 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.0004486274265218526, + "learning_rate": 3.141953535845912e-05, + "loss": 1.0177070180361625e-05, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 355, + "tokens/total": 10744464, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 162318 + }, + { + "epoch": 1.390625, + "grad_norm": 0.01730065606534481, + "learning_rate": 3.11727834939056e-05, + "loss": 7.366175123024732e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00007, + "step": 356, + "tokens/total": 10774976, + "tokens/train_per_sec_per_gpu": 31.71, + "tokens/trainable": 162789 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.4219278395175934, + "learning_rate": 3.092702317708967e-05, + "loss": 0.006937004160135984, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00696, + "step": 357, + "tokens/total": 10805600, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 163254 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.008418221957981586, + "learning_rate": 3.0682264635101276e-05, + "loss": 5.021447941544466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 358, + "tokens/total": 10835920, + "tokens/train_per_sec_per_gpu": 34.74, + "tokens/trainable": 163715 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.3869379460811615, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.0042590186931192875, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00427, + "step": 359, + "tokens/total": 10866048, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 164197 + }, + { + "epoch": 1.40625, + "grad_norm": 0.2765217125415802, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.0020075999200344086, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00201, + "step": 360, + "tokens/total": 10896288, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 164661 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.2115616798400879, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.0020617658738046885, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00206, + "step": 361, + "tokens/total": 10926816, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 165111 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.011683800257742405, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.0001429952826583758, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.31, + "memory/max_allocated (GiB)": 33.31, + "ppl": 1.00014, + "step": 362, + "tokens/total": 10954928, + "tokens/train_per_sec_per_gpu": 30.67, + "tokens/trainable": 165552 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.0025901086628437042, + "learning_rate": 2.9473853553877484e-05, + "loss": 2.3585678718518466e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 363, + "tokens/total": 10985328, + "tokens/train_per_sec_per_gpu": 35.54, + "tokens/trainable": 166026 + }, + { + "epoch": 1.421875, + "grad_norm": 0.14587508141994476, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0011214152909815311, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00112, + "step": 364, + "tokens/total": 11015664, + "tokens/train_per_sec_per_gpu": 38.23, + "tokens/trainable": 166486 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.021550685167312622, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.00018227534019388258, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00018, + "step": 365, + "tokens/total": 11046256, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 166959 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.002367217792198062, + "learning_rate": 2.8761473491751258e-05, + "loss": 2.783137279038783e-05, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00003, + "step": 366, + "tokens/total": 11076288, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 167394 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.02959338203072548, + "learning_rate": 2.8526184124692883e-05, + "loss": 0.0001235282834386453, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00012, + "step": 367, + "tokens/total": 11106416, + "tokens/train_per_sec_per_gpu": 28.63, + "tokens/trainable": 167832 + }, + { + "epoch": 1.4375, + "grad_norm": 0.04836918041110039, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0002190673112636432, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00022, + "step": 368, + "tokens/total": 11136848, + "tokens/train_per_sec_per_gpu": 31.23, + "tokens/trainable": 168242 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.17935794591903687, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.001091235550120473, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00109, + "step": 369, + "tokens/total": 11167376, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168715 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.17555589973926544, + "learning_rate": 2.782696506053033e-05, + "loss": 0.0008262722985818982, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00083, + "step": 370, + "tokens/total": 11197776, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 169147 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.015604093670845032, + "learning_rate": 2.7596140715257824e-05, + "loss": 7.283307058969513e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00007, + "step": 371, + "tokens/total": 11227840, + "tokens/train_per_sec_per_gpu": 34.57, + "tokens/trainable": 169628 + }, + { + "epoch": 1.453125, + "grad_norm": 0.001744552399031818, + "learning_rate": 2.7366456756428184e-05, + "loss": 1.2661729670071509e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00001, + "step": 372, + "tokens/total": 11258176, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 170085 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.03369157761335373, + "learning_rate": 2.7137922742128486e-05, + "loss": 0.00013295823009684682, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00013, + "step": 373, + "tokens/total": 11288336, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 170584 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.11042775958776474, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00030668824911117554, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00031, + "step": 374, + "tokens/total": 11318848, + "tokens/train_per_sec_per_gpu": 34.92, + "tokens/trainable": 171058 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.015241619199514389, + "learning_rate": 2.6684342539801933e-05, + "loss": 5.481348853209056e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 375, + "tokens/total": 11349168, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 171518 + }, + { + "epoch": 1.46875, + "grad_norm": 0.07663743942975998, + "learning_rate": 2.645931522709877e-05, + "loss": 0.0004268307238817215, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00043, + "step": 376, + "tokens/total": 11379328, + "tokens/train_per_sec_per_gpu": 40.72, + "tokens/trainable": 172007 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.3878970146179199, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.006505412515252829, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00653, + "step": 377, + "tokens/total": 11409568, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 172468 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.011625024490058422, + "learning_rate": 2.6012832999747916e-05, + "loss": 7.441644265782088e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00007, + "step": 378, + "tokens/total": 11439968, + "tokens/train_per_sec_per_gpu": 33.97, + "tokens/trainable": 172923 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.8276505470275879, + "learning_rate": 2.579139666504821e-05, + "loss": 0.022178810089826584, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02243, + "step": 379, + "tokens/total": 11470496, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 173370 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0010690620401874185, + "learning_rate": 2.557117581955798e-05, + "loss": 1.1797061233664863e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00001, + "step": 380, + "tokens/total": 11500720, + "tokens/train_per_sec_per_gpu": 34.68, + "tokens/trainable": 173837 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.06872054934501648, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.0004525431140791625, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00045, + "step": 381, + "tokens/total": 11531280, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 174294 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.025393052026629448, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0001451169518986717, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 382, + "tokens/total": 11561264, + "tokens/train_per_sec_per_gpu": 31.04, + "tokens/trainable": 174701 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.006796801462769508, + "learning_rate": 2.491789760603361e-05, + "loss": 9.397654503118247e-05, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 383, + "tokens/total": 11591616, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 175131 + }, + { + "epoch": 1.5, + "grad_norm": 0.35059845447540283, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.004584255628287792, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00459, + "step": 384, + "tokens/total": 11622000, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 175586 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.33106186985969543, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.010121386498212814, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01017, + "step": 385, + "tokens/total": 11652352, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 176026 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.003417555009946227, + "learning_rate": 2.427588563158384e-05, + "loss": 4.207541496725753e-05, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 386, + "tokens/total": 11682736, + "tokens/train_per_sec_per_gpu": 34.05, + "tokens/trainable": 176512 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.24088126420974731, + "learning_rate": 2.406442693028651e-05, + "loss": 0.004325853195041418, + "memory/device_reserved (GiB)": 34.98, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00434, + "step": 387, + "tokens/total": 11713136, + "tokens/train_per_sec_per_gpu": 30.64, + "tokens/trainable": 176943 + }, + { + "epoch": 1.515625, + "grad_norm": 0.04677414149045944, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00039886703598313034, + "memory/device_reserved (GiB)": 35.23, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.0004, + "step": 388, + "tokens/total": 11743808, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 177370 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.01845633052289486, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00019935239106416702, + "memory/device_reserved (GiB)": 35.58, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0002, + "step": 389, + "tokens/total": 11774144, + "tokens/train_per_sec_per_gpu": 35.42, + "tokens/trainable": 177851 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.19680730998516083, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0020088059827685356, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00201, + "step": 390, + "tokens/total": 11804272, + "tokens/train_per_sec_per_gpu": 35.03, + "tokens/trainable": 178310 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.029754316434264183, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.00013066174869891256, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00013, + "step": 391, + "tokens/total": 11832432, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 178736 + }, + { + "epoch": 1.53125, + "grad_norm": 0.47083696722984314, + "learning_rate": 2.3026617866382657e-05, + "loss": 0.006020379718393087, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00604, + "step": 392, + "tokens/total": 11862960, + "tokens/train_per_sec_per_gpu": 36.57, + "tokens/trainable": 179233 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.02095848135650158, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023951532784849405, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893232, + "tokens/train_per_sec_per_gpu": 38.68, + "tokens/trainable": 179730 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.09784477949142456, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0008653341792523861, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00087, + "step": 394, + "tokens/total": 11923664, + "tokens/train_per_sec_per_gpu": 37.87, + "tokens/trainable": 180227 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.016663849353790283, + "learning_rate": 2.2419829946830123e-05, + "loss": 0.0002501910785213113, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00025, + "step": 395, + "tokens/total": 11954032, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 180687 + }, + { + "epoch": 1.546875, + "grad_norm": 0.06284014135599136, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0006274969782680273, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00063, + "step": 396, + "tokens/total": 11984512, + "tokens/train_per_sec_per_gpu": 33.05, + "tokens/trainable": 181141 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.06682170182466507, + "learning_rate": 2.202206960760984e-05, + "loss": 0.0008653995464555919, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00087, + "step": 397, + "tokens/total": 12014880, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 181648 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.048094261437654495, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0006572521524503827, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00066, + "step": 398, + "tokens/total": 12044928, + "tokens/train_per_sec_per_gpu": 34.17, + "tokens/trainable": 182109 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.288680762052536, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.0029174773953855038, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00292, + "step": 399, + "tokens/total": 12075344, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 182559 + }, + { + "epoch": 1.5625, + "grad_norm": 0.009714324958622456, + "learning_rate": 2.1435742029681725e-05, + "loss": 8.201875607483089e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105616, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 182985 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.00939064659178257, + "learning_rate": 2.124308220868431e-05, + "loss": 0.00012157741002738476, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00012, + "step": 401, + "tokens/total": 12136080, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 183440 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.035272955894470215, + "learning_rate": 2.105182715082638e-05, + "loss": 0.0004544599214568734, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00045, + "step": 402, + "tokens/total": 12166240, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 183917 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.22718806564807892, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0024650408886373043, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00247, + "step": 403, + "tokens/total": 12196544, + "tokens/train_per_sec_per_gpu": 30.51, + "tokens/trainable": 184348 + }, + { + "epoch": 1.578125, + "grad_norm": 0.013060114346444607, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.0001124740592786111, + "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": 31.69, + "tokens/trainable": 184786 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.052736442536115646, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0003598350449465215, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00036, + "step": 405, + "tokens/total": 12257264, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 185249 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.00711337523534894, + "learning_rate": 2.0301012845054956e-05, + "loss": 8.437763608526438e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 406, + "tokens/total": 12287584, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185680 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.00903877429664135, + "learning_rate": 2.011689980574966e-05, + "loss": 0.00010423596540931612, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0001, + "step": 407, + "tokens/total": 12317984, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 186100 + }, + { + "epoch": 1.59375, + "grad_norm": 0.6410778164863586, + "learning_rate": 1.993423839463052e-05, + "loss": 0.004257278051227331, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00427, + "step": 408, + "tokens/total": 12348144, + "tokens/train_per_sec_per_gpu": 40.16, + "tokens/trainable": 186565 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.013556886464357376, + "learning_rate": 1.975303621298445e-05, + "loss": 0.0001645060838200152, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00016, + "step": 409, + "tokens/total": 12378544, + "tokens/train_per_sec_per_gpu": 36.06, + "tokens/trainable": 187017 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.10837272554636002, + "learning_rate": 1.957330080137385e-05, + "loss": 0.0001806107466109097, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00018, + "step": 410, + "tokens/total": 12408784, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 187468 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.005454800557345152, + "learning_rate": 1.9395039639322864e-05, + "loss": 9.649700950831175e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0001, + "step": 411, + "tokens/total": 12438832, + "tokens/train_per_sec_per_gpu": 37.94, + "tokens/trainable": 187961 + }, + { + "epoch": 1.609375, + "grad_norm": 0.003915195353329182, + "learning_rate": 1.9218260145006073e-05, + "loss": 5.392331149778329e-05, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 412, + "tokens/total": 12469296, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 188447 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.024293167516589165, + "learning_rate": 1.904296967493982e-05, + "loss": 0.00030749564757570624, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00031, + "step": 413, + "tokens/total": 12499392, + "tokens/train_per_sec_per_gpu": 36.19, + "tokens/trainable": 188903 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.1073823794722557, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.0014971208292990923, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.0015, + "step": 414, + "tokens/total": 12529952, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 189398 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.025999421253800392, + "learning_rate": 1.869688492349885e-05, + "loss": 0.00022489691036753356, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 415, + "tokens/total": 12560240, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 189850 + }, + { + "epoch": 1.625, + "grad_norm": 0.033106278628110886, + "learning_rate": 1.85261050441233e-05, + "loss": 0.0003824468585662544, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00038, + "step": 416, + "tokens/total": 12590736, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 190306 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.010407397523522377, + "learning_rate": 1.8356842992397304e-05, + "loss": 8.543908188585192e-05, + "memory/device_reserved (GiB)": 38.05, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00009, + "step": 417, + "tokens/total": 12621168, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 190746 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.007676505949348211, + "learning_rate": 1.8189105812005714e-05, + "loss": 9.762673289515078e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.0001, + "step": 418, + "tokens/total": 12651456, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 191156 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.008882309310138226, + "learning_rate": 1.802290048317732e-05, + "loss": 0.00011588008783292025, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00012, + "step": 419, + "tokens/total": 12681760, + "tokens/train_per_sec_per_gpu": 29.78, + "tokens/trainable": 191573 + }, + { + "epoch": 1.640625, + "grad_norm": 0.01079952996224165, + "learning_rate": 1.785823392239424e-05, + "loss": 0.00015772700135130435, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00016, + "step": 420, + "tokens/total": 12712144, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 192014 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.10464993119239807, + "learning_rate": 1.7695112982104225e-05, + "loss": 0.0007448438555002213, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00075, + "step": 421, + "tokens/total": 12742400, + "tokens/train_per_sec_per_gpu": 33.16, + "tokens/trainable": 192453 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.033524010330438614, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0003343333082739264, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00033, + "step": 422, + "tokens/total": 12772288, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 192886 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.0037654207553714514, + "learning_rate": 1.7373535050913946e-05, + "loss": 4.5559077989310026e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00005, + "step": 423, + "tokens/total": 12802880, + "tokens/train_per_sec_per_gpu": 30.19, + "tokens/trainable": 193323 + }, + { + "epoch": 1.65625, + "grad_norm": 0.06157264485955238, + "learning_rate": 1.721509144218405e-05, + "loss": 0.0005529897753149271, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00055, + "step": 424, + "tokens/total": 12833456, + "tokens/train_per_sec_per_gpu": 32.84, + "tokens/trainable": 193753 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.009139563888311386, + "learning_rate": 1.705822021773101e-05, + "loss": 0.00013727511395700276, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00014, + "step": 425, + "tokens/total": 12863952, + "tokens/train_per_sec_per_gpu": 36.43, + "tokens/trainable": 194240 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.32006195187568665, + "learning_rate": 1.69029279056068e-05, + "loss": 0.005133518483489752, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00515, + "step": 426, + "tokens/total": 12894352, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 194680 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.19554191827774048, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.0020202554296702147, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00202, + "step": 427, + "tokens/total": 12924688, + "tokens/train_per_sec_per_gpu": 39.09, + "tokens/trainable": 195186 + }, + { + "epoch": 1.671875, + "grad_norm": 0.010584031231701374, + "learning_rate": 1.659710580175893e-05, + "loss": 7.440667104674503e-05, + "memory/device_reserved (GiB)": 36.05, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00007, + "step": 428, + "tokens/total": 12955072, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 195693 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.005364356562495232, + "learning_rate": 1.644658873654133e-05, + "loss": 6.896692502778023e-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": 37.33, + "tokens/trainable": 196130 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.0026337867602705956, + "learning_rate": 1.629767603613508e-05, + "loss": 5.40713153895922e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00005, + "step": 430, + "tokens/total": 13015600, + "tokens/train_per_sec_per_gpu": 29.85, + "tokens/trainable": 196530 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.08371307700872421, + "learning_rate": 1.615037389740547e-05, + "loss": 0.000647605920676142, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00065, + "step": 431, + "tokens/total": 13045728, + "tokens/train_per_sec_per_gpu": 30.2, + "tokens/trainable": 196974 + }, + { + "epoch": 1.6875, + "grad_norm": 0.12106083333492279, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0010971315205097198, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0011, + "step": 432, + "tokens/total": 13075920, + "tokens/train_per_sec_per_gpu": 35.55, + "tokens/trainable": 197454 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.0016024125507101417, + "learning_rate": 1.5860625757072092e-05, + "loss": 2.5408877263544127e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00003, + "step": 433, + "tokens/total": 13106128, + "tokens/train_per_sec_per_gpu": 31.68, + "tokens/trainable": 197902 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.006628716830164194, + "learning_rate": 1.571819181307116e-05, + "loss": 9.017687989398837e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00009, + "step": 434, + "tokens/total": 13136368, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 198374 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.023524144664406776, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0003021995653398335, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0003, + "step": 435, + "tokens/total": 13166704, + "tokens/train_per_sec_per_gpu": 37.86, + "tokens/trainable": 198846 + }, + { + "epoch": 1.703125, + "grad_norm": 0.1354794055223465, + "learning_rate": 1.543823381344311e-05, + "loss": 0.001777412835508585, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00178, + "step": 436, + "tokens/total": 13197344, + "tokens/train_per_sec_per_gpu": 38.06, + "tokens/trainable": 199320 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.0022158098872750998, + "learning_rate": 1.5300721408011114e-05, + "loss": 2.7318394131725654e-05, + "memory/device_reserved (GiB)": 36.06, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 437, + "tokens/total": 13227664, + "tokens/train_per_sec_per_gpu": 37.77, + "tokens/trainable": 199777 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.005856889300048351, + "learning_rate": 1.5164861051607254e-05, + "loss": 7.72368221078068e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00008, + "step": 438, + "tokens/total": 13257632, + "tokens/train_per_sec_per_gpu": 31.16, + "tokens/trainable": 200194 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.007335511967539787, + "learning_rate": 1.5030658397935521e-05, + "loss": 7.384424679912627e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 439, + "tokens/total": 13288240, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 200663 + }, + { + "epoch": 1.71875, + "grad_norm": 0.00513842236250639, + "learning_rate": 1.4898119031716104e-05, + "loss": 8.217831782530993e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00008, + "step": 440, + "tokens/total": 13318992, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 201121 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.0026071977335959673, + "learning_rate": 1.476724846845306e-05, + "loss": 3.909884981112555e-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": 40.13, + "tokens/trainable": 201598 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.0043723853304982185, + "learning_rate": 1.463805215420471e-05, + "loss": 4.3211588490521535e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 442, + "tokens/total": 13379648, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 202068 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.032961804419755936, + "learning_rate": 1.451053546535705e-05, + "loss": 0.00024515652330592275, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00025, + "step": 443, + "tokens/total": 13409936, + "tokens/train_per_sec_per_gpu": 35.82, + "tokens/trainable": 202524 + }, + { + "epoch": 1.734375, + "grad_norm": 0.12496478110551834, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0008977800025604665, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.0009, + "step": 444, + "tokens/total": 13440288, + "tokens/train_per_sec_per_gpu": 36.64, + "tokens/trainable": 202996 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.016452984884381294, + "learning_rate": 1.4260562119706606e-05, + "loss": 0.00011743961658794433, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00012, + "step": 445, + "tokens/total": 13468880, + "tokens/train_per_sec_per_gpu": 39.58, + "tokens/trainable": 203470 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.009045973420143127, + "learning_rate": 1.413811586531508e-05, + "loss": 8.700705802766606e-05, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00009, + "step": 446, + "tokens/total": 13499184, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 203919 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.038517843931913376, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.0003440188302192837, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00034, + "step": 447, + "tokens/total": 13529632, + "tokens/train_per_sec_per_gpu": 33.61, + "tokens/trainable": 204376 + }, + { + "epoch": 1.75, + "grad_norm": 0.019700370728969574, + "learning_rate": 1.3898329670629645e-05, + "loss": 0.00021027779439464211, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00021, + "step": 448, + "tokens/total": 13560080, + "tokens/train_per_sec_per_gpu": 32.3, + "tokens/trainable": 204821 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.0006826155004091561, + "learning_rate": 1.3780999708818058e-05, + "loss": 1.4103720786806662e-05, + "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": 33.92, + "tokens/trainable": 205271 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.008357573300600052, + "learning_rate": 1.3665385037857758e-05, + "loss": 9.161404886981472e-05, + "memory/device_reserved (GiB)": 34.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00009, + "step": 450, + "tokens/total": 13620960, + "tokens/train_per_sec_per_gpu": 31.24, + "tokens/trainable": 205723 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.01647551730275154, + "learning_rate": 1.3551490468947126e-05, + "loss": 0.00017527706222608685, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00018, + "step": 451, + "tokens/total": 13651344, + "tokens/train_per_sec_per_gpu": 32.75, + "tokens/trainable": 206169 + }, + { + "epoch": 1.765625, + "grad_norm": 0.03474664315581322, + "learning_rate": 1.3439320741704075e-05, + "loss": 0.0004909664858132601, + "memory/device_reserved (GiB)": 35.04, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00049, + "step": 452, + "tokens/total": 13681792, + "tokens/train_per_sec_per_gpu": 33.01, + "tokens/trainable": 206592 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.005533512681722641, + "learning_rate": 1.3328880523968808e-05, + "loss": 8.479113603243604e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00008, + "step": 453, + "tokens/total": 13712256, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 207055 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.028869982808828354, + "learning_rate": 1.3220174411609587e-05, + "loss": 7.264575106091797e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.00007, + "step": 454, + "tokens/total": 13742128, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 207494 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.0076578096486628056, + "learning_rate": 1.3113206928331471e-05, + "loss": 2.4288401618832722e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 455, + "tokens/total": 13772160, + "tokens/train_per_sec_per_gpu": 33.22, + "tokens/trainable": 207956 + }, + { + "epoch": 1.78125, + "grad_norm": 0.0006251306622289121, + "learning_rate": 1.300798252548806e-05, + "loss": 1.4033696061233059e-05, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00001, + "step": 456, + "tokens/total": 13802448, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 208408 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.0268402099609375, + "learning_rate": 1.2904505581896265e-05, + "loss": 0.00014731910778209567, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00015, + "step": 457, + "tokens/total": 13832880, + "tokens/train_per_sec_per_gpu": 36.37, + "tokens/trainable": 208889 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.060711171478033066, + "learning_rate": 1.2802780403654082e-05, + "loss": 0.00020443923131097108, + "memory/device_reserved (GiB)": 35.64, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.0002, + "step": 458, + "tokens/total": 13862880, + "tokens/train_per_sec_per_gpu": 34.44, + "tokens/trainable": 209336 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.12638092041015625, + "learning_rate": 1.2702811223961408e-05, + "loss": 0.0009473193786107004, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00095, + "step": 459, + "tokens/total": 13893440, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 209797 + }, + { + "epoch": 1.796875, + "grad_norm": 0.0015253991587087512, + "learning_rate": 1.2604602202943861e-05, + "loss": 2.2379259462468326e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00002, + "step": 460, + "tokens/total": 13923824, + "tokens/train_per_sec_per_gpu": 36.49, + "tokens/trainable": 210240 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.013747433200478554, + "learning_rate": 1.2508157427479686e-05, + "loss": 9.997214510804042e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0001, + "step": 461, + "tokens/total": 13954208, + "tokens/train_per_sec_per_gpu": 35.22, + "tokens/trainable": 210702 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.08502306789159775, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.0007195118232630193, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00072, + "step": 462, + "tokens/total": 13984784, + "tokens/train_per_sec_per_gpu": 31.78, + "tokens/trainable": 211140 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.016438405960798264, + "learning_rate": 1.2320576593470082e-05, + "loss": 9.172871796181425e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00009, + "step": 463, + "tokens/total": 14014896, + "tokens/train_per_sec_per_gpu": 32.93, + "tokens/trainable": 211575 + }, + { + "epoch": 1.8125, + "grad_norm": 0.09976232051849365, + "learning_rate": 1.2229448340928828e-05, + "loss": 0.000811650650575757, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00081, + "step": 464, + "tokens/total": 14045312, + "tokens/train_per_sec_per_gpu": 39.32, + "tokens/trainable": 212083 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.00435422919690609, + "learning_rate": 1.2140099945624458e-05, + "loss": 2.5664587155915797e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00003, + "step": 465, + "tokens/total": 14075840, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 212558 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.0026768327224999666, + "learning_rate": 1.205253512570841e-05, + "loss": 3.4629334550118074e-05, + "memory/device_reserved (GiB)": 35.97, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 466, + "tokens/total": 14106016, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 213014 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0026948358863592148, + "learning_rate": 1.1966757525110255e-05, + "loss": 4.344901753938757e-05, + "memory/device_reserved (GiB)": 37.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00004, + "step": 467, + "tokens/total": 14136512, + "tokens/train_per_sec_per_gpu": 34.3, + "tokens/trainable": 213465 + }, + { + "epoch": 1.828125, + "grad_norm": 0.006161310710012913, + "learning_rate": 1.1882770713386095e-05, + "loss": 4.4334883568808436e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 468, + "tokens/total": 14166832, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 213940 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.010646386072039604, + "learning_rate": 1.180057818556998e-05, + "loss": 2.2965485186432488e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00002, + "step": 469, + "tokens/total": 14197360, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 214415 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.0007588419830426574, + "learning_rate": 1.1720183362028494e-05, + "loss": 1.3413812666840386e-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": 34.61, + "tokens/trainable": 214880 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.017162201926112175, + "learning_rate": 1.1641589588318387e-05, + "loss": 0.0001371123653370887, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00014, + "step": 471, + "tokens/total": 14257888, + "tokens/train_per_sec_per_gpu": 33.74, + "tokens/trainable": 215334 + }, + { + "epoch": 1.84375, + "grad_norm": 0.0008721952908672392, + "learning_rate": 1.1564800135047418e-05, + "loss": 1.1102063581347466e-05, + "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": 35.27, + "tokens/trainable": 215822 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.029533211141824722, + "learning_rate": 1.148981819773816e-05, + "loss": 0.0002879542880691588, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00029, + "step": 473, + "tokens/total": 14318448, + "tokens/train_per_sec_per_gpu": 31.18, + "tokens/trainable": 216239 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.0012399809202179313, + "learning_rate": 1.1416646896695086e-05, + "loss": 1.8869191990233958e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00002, + "step": 474, + "tokens/total": 14348912, + "tokens/train_per_sec_per_gpu": 33.67, + "tokens/trainable": 216691 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.22101251780986786, + "learning_rate": 1.1345289276874717e-05, + "loss": 0.0024789171293377876, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00248, + "step": 475, + "tokens/total": 14379440, + "tokens/train_per_sec_per_gpu": 33.41, + "tokens/trainable": 217162 + }, + { + "epoch": 1.859375, + "grad_norm": 0.004043657332658768, + "learning_rate": 1.1275748307758873e-05, + "loss": 4.4153628550702706e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 476, + "tokens/total": 14409840, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 217638 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.022147774696350098, + "learning_rate": 1.1208026883231147e-05, + "loss": 0.00015884214371908456, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00016, + "step": 477, + "tokens/total": 14440256, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 218053 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.0010556846391409636, + "learning_rate": 1.1142127821456433e-05, + "loss": 1.8117327272193506e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 478, + "tokens/total": 14470704, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 218503 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.019303888082504272, + "learning_rate": 1.1078053864763674e-05, + "loss": 0.0001503409876022488, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00015, + "step": 479, + "tokens/total": 14501072, + "tokens/train_per_sec_per_gpu": 36.47, + "tokens/trainable": 218964 + }, + { + "epoch": 1.875, + "grad_norm": 0.008807498030364513, + "learning_rate": 1.1015807679531756e-05, + "loss": 6.010168726788834e-05, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00006, + "step": 480, + "tokens/total": 14531200, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 219412 + }, + { + "epoch": 1.87890625, + "grad_norm": 0.0371716283261776, + "learning_rate": 1.0955391856078528e-05, + "loss": 0.00035582270356826484, + "memory/device_reserved (GiB)": 38.63, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00036, + "step": 481, + "tokens/total": 14561152, + "tokens/train_per_sec_per_gpu": 38.62, + "tokens/trainable": 219886 + }, + { + "epoch": 1.8828125, + "grad_norm": 0.0007208790630102158, + "learning_rate": 1.0896808908553007e-05, + "loss": 1.2405967027007136e-05, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 482, + "tokens/total": 14591264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 220343 + }, + { + "epoch": 1.88671875, + "grad_norm": 0.019832292571663857, + "learning_rate": 1.0840061274830763e-05, + "loss": 0.0002267304516863078, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00023, + "step": 483, + "tokens/total": 14621568, + "tokens/train_per_sec_per_gpu": 35.53, + "tokens/trainable": 220825 + }, + { + "epoch": 1.890625, + "grad_norm": 0.0017411233857274055, + "learning_rate": 1.0785151316412473e-05, + "loss": 2.01847760763485e-05, + "memory/device_reserved (GiB)": 37.77, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 484, + "tokens/total": 14652064, + "tokens/train_per_sec_per_gpu": 32.83, + "tokens/trainable": 221282 + }, + { + "epoch": 1.89453125, + "grad_norm": 0.0007287322077900171, + "learning_rate": 1.0732081318325639e-05, + "loss": 9.605650120647624e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 485, + "tokens/total": 14682608, + "tokens/train_per_sec_per_gpu": 38.26, + "tokens/trainable": 221780 + }, + { + "epoch": 1.8984375, + "grad_norm": 0.0007819020538590848, + "learning_rate": 1.0680853489029501e-05, + "loss": 1.6115096514113247e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 486, + "tokens/total": 14713168, + "tokens/train_per_sec_per_gpu": 31.89, + "tokens/trainable": 222210 + }, + { + "epoch": 1.90234375, + "grad_norm": 0.014062542468309402, + "learning_rate": 1.0631469960323152e-05, + "loss": 0.00013912066060584038, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00014, + "step": 487, + "tokens/total": 14743536, + "tokens/train_per_sec_per_gpu": 29.67, + "tokens/trainable": 222646 + }, + { + "epoch": 1.90625, + "grad_norm": 0.13683916628360748, + "learning_rate": 1.0583932787256783e-05, + "loss": 0.0015979098388925195, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.0016, + "step": 488, + "tokens/total": 14773728, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 223066 + }, + { + "epoch": 1.91015625, + "grad_norm": 0.5263968706130981, + "learning_rate": 1.0538243948046206e-05, + "loss": 0.00542115792632103, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00544, + "step": 489, + "tokens/total": 14803744, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 223512 + }, + { + "epoch": 1.9140625, + "grad_norm": 0.03019803576171398, + "learning_rate": 1.0494405343990523e-05, + "loss": 0.0002152348606614396, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00022, + "step": 490, + "tokens/total": 14832112, + "tokens/train_per_sec_per_gpu": 39.3, + "tokens/trainable": 223956 + }, + { + "epoch": 1.91796875, + "grad_norm": 0.047021783888339996, + "learning_rate": 1.0452418799392985e-05, + "loss": 0.00024185153597500175, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.00024, + "step": 491, + "tokens/total": 14860560, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 224416 + }, + { + "epoch": 1.921875, + "grad_norm": 0.0029584276489913464, + "learning_rate": 1.0412286061485102e-05, + "loss": 3.368399484315887e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 492, + "tokens/total": 14890848, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 224883 + }, + { + "epoch": 1.92578125, + "grad_norm": 0.005298682954162359, + "learning_rate": 1.03740088003539e-05, + "loss": 3.912312240572646e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 493, + "tokens/total": 14921008, + "tokens/train_per_sec_per_gpu": 35.8, + "tokens/trainable": 225324 + }, + { + "epoch": 1.9296875, + "grad_norm": 0.007532566320151091, + "learning_rate": 1.0337588608872463e-05, + "loss": 7.047319377306849e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00007, + "step": 494, + "tokens/total": 14951360, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 225783 + }, + { + "epoch": 1.93359375, + "grad_norm": 0.006395815405994654, + "learning_rate": 1.0303027002633622e-05, + "loss": 4.224019357934594e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 495, + "tokens/total": 14981552, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 226281 + }, + { + "epoch": 1.9375, + "grad_norm": 0.0026577117387205362, + "learning_rate": 1.0270325419886884e-05, + "loss": 1.877904105640482e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 496, + "tokens/total": 15011584, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 226735 + }, + { + "epoch": 1.94140625, + "grad_norm": 0.003066121833398938, + "learning_rate": 1.0239485221478599e-05, + "loss": 3.809008558164351e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00004, + "step": 497, + "tokens/total": 15042016, + "tokens/train_per_sec_per_gpu": 36.66, + "tokens/trainable": 227198 + }, + { + "epoch": 1.9453125, + "grad_norm": 0.0003162898647133261, + "learning_rate": 1.0210507690795292e-05, + "loss": 9.668269740359392e-06, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.68, + "memory/max_allocated (GiB)": 33.68, + "ppl": 1.00001, + "step": 498, + "tokens/total": 15071984, + "tokens/train_per_sec_per_gpu": 31.4, + "tokens/trainable": 227623 + }, + { + "epoch": 1.94921875, + "grad_norm": 0.035544853657484055, + "learning_rate": 1.0183394033710305e-05, + "loss": 0.00024893757654353976, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00025, + "step": 499, + "tokens/total": 15102192, + "tokens/train_per_sec_per_gpu": 32.48, + "tokens/trainable": 228074 + }, + { + "epoch": 1.953125, + "grad_norm": 0.005648774094879627, + "learning_rate": 1.0158145378533583e-05, + "loss": 4.911683936370537e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 500, + "tokens/total": 15132560, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 228541 + }, + { + "epoch": 1.95703125, + "grad_norm": 0.0024769131559878588, + "learning_rate": 1.0134762775964726e-05, + "loss": 2.570717151684221e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00003, + "step": 501, + "tokens/total": 15163152, + "tokens/train_per_sec_per_gpu": 38.07, + "tokens/trainable": 229031 + }, + { + "epoch": 1.9609375, + "grad_norm": 0.005817459896206856, + "learning_rate": 1.0113247199049278e-05, + "loss": 3.874715912388638e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 502, + "tokens/total": 15193456, + "tokens/train_per_sec_per_gpu": 39.01, + "tokens/trainable": 229512 + }, + { + "epoch": 1.96484375, + "grad_norm": 0.213705375790596, + "learning_rate": 1.0093599543138205e-05, + "loss": 0.0010556308552622795, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00106, + "step": 503, + "tokens/total": 15223680, + "tokens/train_per_sec_per_gpu": 34.53, + "tokens/trainable": 229989 + }, + { + "epoch": 1.96875, + "grad_norm": 0.0057237111032009125, + "learning_rate": 1.0075820625850675e-05, + "loss": 2.9780632758047432e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00003, + "step": 504, + "tokens/total": 15254336, + "tokens/train_per_sec_per_gpu": 36.15, + "tokens/trainable": 230447 + }, + { + "epoch": 1.97265625, + "grad_norm": 0.017064962536096573, + "learning_rate": 1.0059911187040013e-05, + "loss": 0.0001428169198334217, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00014, + "step": 505, + "tokens/total": 15284496, + "tokens/train_per_sec_per_gpu": 34.6, + "tokens/trainable": 230879 + }, + { + "epoch": 1.9765625, + "grad_norm": 0.004507618024945259, + "learning_rate": 1.0045871888762893e-05, + "loss": 1.688709744485095e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00002, + "step": 506, + "tokens/total": 15314976, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 231348 + }, + { + "epoch": 1.98046875, + "grad_norm": 0.0015678065828979015, + "learning_rate": 1.003370331525184e-05, + "loss": 2.9251012165332213e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00003, + "step": 507, + "tokens/total": 15345280, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 231819 + }, + { + "epoch": 1.984375, + "grad_norm": 0.0032501902896910906, + "learning_rate": 1.002340597289085e-05, + "loss": 4.6847708290442824e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 508, + "tokens/total": 15375424, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 232269 + }, + { + "epoch": 1.98828125, + "grad_norm": 0.003683724207803607, + "learning_rate": 1.0014980290194387e-05, + "loss": 3.591051427065395e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 509, + "tokens/total": 15405872, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 232716 + }, + { + "epoch": 1.9921875, + "grad_norm": 0.0030913210939615965, + "learning_rate": 1.0008426617789489e-05, + "loss": 2.7738915377995e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 510, + "tokens/total": 15436128, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 233178 + }, + { + "epoch": 1.99609375, + "grad_norm": 0.04374156519770622, + "learning_rate": 1.0003745228401215e-05, + "loss": 4.1764309571590275e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00004, + "step": 511, + "tokens/total": 15466464, + "tokens/train_per_sec_per_gpu": 36.21, + "tokens/trainable": 233647 + }, + { + "epoch": 2.0, + "grad_norm": 0.00487576425075531, + "learning_rate": 1.0000936316841296e-05, + "loss": 3.3267013350268826e-05, + "memory/device_reserved (GiB)": 38.44, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00003, + "step": 512, + "tokens/total": 15496368, + "tokens/train_per_sec_per_gpu": 29.68, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..089deefa08c5d5306c8578653e481502a39478e5 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1dfff037f358f741e8e719b2d53bad6b23c141b9ee9e73007acfd221d44d9ee6 +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..bfbb6cdd7b40cffcca231364b12f201277db20e4 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8d5dcd0c70cfdf223d81701e18715a6127ba78a5b01df88c26fe7b96256a4edd +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..9ff7e35102f980691267d1358fa51e401677a704 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..34b7fc400f1006f136b1c89d6c58cf3d17830d72 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..07d38539bf6c40115a3406c170b5add26e76769d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..78025d53a11d9769f5bef48e19240d25263b017d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/README.md b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/adapter_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ce2ecb2cf76a5f2877ccc6964207b6a3851a10f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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", + "down_proj", + "k_proj", + "gate_proj", + "up_proj", + "v_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f17d572f54e3ee22f92b9fbc36dff73f4ff6ef0f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:278cb04bb02f53eadd1cf3b2e7577eb26c28da869e33ca30f5e004fc0c7079ba +size 547777976 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/chat_template.jinja b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/optimizer.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..13a5d11cd0812c8034521137d71c9e245baa6a0d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:054b55ef7f39f69c500d2a27bf61c265a337065d51e5beb28cb98aba196d15ba +size 1048106435 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/rng_state.pth b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..4ac2a4558d28c1b7802561172e30decf438ff206 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/scheduler.pt b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..433f82b14836de77d366e2c961bc5bdacdc9eb63 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokenizer.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokenizer_config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokens_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..391bb3d32a2053c8ef21df09dfdc71aecacae5ae --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/trainer_state.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..95362408b951f4a1b06677b5dbb5de6af2c28b75 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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": 1.1789305210113525, + "learning_rate": 0.0, + "loss": 0.13649286329746246, + "memory/device_reserved (GiB)": 34.94, + "memory/max_active (GiB)": 32.98, + "memory/max_allocated (GiB)": 32.98, + "ppl": 1.14625, + "step": 1, + "tokens/total": 30464, + "tokens/train_per_sec_per_gpu": 30.5, + "tokens/trainable": 470 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.9851033091545105, + "learning_rate": 4.000000000000001e-06, + "loss": 0.12611928582191467, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.13442, + "step": 2, + "tokens/total": 60800, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 922 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9737245440483093, + "learning_rate": 8.000000000000001e-06, + "loss": 0.140651136636734, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.15102, + "step": 3, + "tokens/total": 91136, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 1396 + }, + { + "epoch": 0.015625, + "grad_norm": 0.9809151887893677, + "learning_rate": 1.2e-05, + "loss": 0.12868642807006836, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.13733, + "step": 4, + "tokens/total": 121552, + "tokens/train_per_sec_per_gpu": 38.32, + "tokens/trainable": 1850 + }, + { + "epoch": 0.01953125, + "grad_norm": 0.9113903045654297, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13655595481395721, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.14632, + "step": 5, + "tokens/total": 152304, + "tokens/train_per_sec_per_gpu": 34.99, + "tokens/trainable": 2302 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.1376721858978271, + "learning_rate": 2e-05, + "loss": 0.10808873176574707, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.11415, + "step": 6, + "tokens/total": 182448, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 2742 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.7545889616012573, + "learning_rate": 2.4e-05, + "loss": 0.09906798601150513, + "memory/device_reserved (GiB)": 35.85, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.10414, + "step": 7, + "tokens/total": 212736, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 3208 + }, + { + "epoch": 0.03125, + "grad_norm": 1.3217154741287231, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.08243394643068314, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.08593, + "step": 8, + "tokens/total": 243024, + "tokens/train_per_sec_per_gpu": 31.98, + "tokens/trainable": 3655 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.7662289142608643, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.04795612767338753, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04912, + "step": 9, + "tokens/total": 271264, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 4091 + }, + { + "epoch": 0.0390625, + "grad_norm": 2.8066670894622803, + "learning_rate": 3.6e-05, + "loss": 0.11464612931013107, + "memory/device_reserved (GiB)": 36.07, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.12148, + "step": 10, + "tokens/total": 301520, + "tokens/train_per_sec_per_gpu": 35.21, + "tokens/trainable": 4553 + }, + { + "epoch": 0.04296875, + "grad_norm": 4.6501383781433105, + "learning_rate": 4e-05, + "loss": 0.09800341725349426, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.10297, + "step": 11, + "tokens/total": 331808, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 4992 + }, + { + "epoch": 0.046875, + "grad_norm": 3.8860652446746826, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.10282482206821442, + "memory/device_reserved (GiB)": 36.08, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.1083, + "step": 12, + "tokens/total": 362224, + "tokens/train_per_sec_per_gpu": 38.66, + "tokens/trainable": 5494 + }, + { + "epoch": 0.05078125, + "grad_norm": 4.133975505828857, + "learning_rate": 4.8e-05, + "loss": 0.07831095159053802, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08146, + "step": 13, + "tokens/total": 392432, + "tokens/train_per_sec_per_gpu": 35.3, + "tokens/trainable": 5933 + }, + { + "epoch": 0.0546875, + "grad_norm": 2.6654460430145264, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.11425333470106125, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.12104, + "step": 14, + "tokens/total": 422784, + "tokens/train_per_sec_per_gpu": 30.01, + "tokens/trainable": 6369 + }, + { + "epoch": 0.05859375, + "grad_norm": 3.4556033611297607, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09427995979785919, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.09887, + "step": 15, + "tokens/total": 453376, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 6820 + }, + { + "epoch": 0.0625, + "grad_norm": 2.517883062362671, + "learning_rate": 6e-05, + "loss": 0.07633613049983978, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.07933, + "step": 16, + "tokens/total": 483504, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 7258 + }, + { + "epoch": 0.06640625, + "grad_norm": 5.47124719619751, + "learning_rate": 6.400000000000001e-05, + "loss": 0.07268983870744705, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0754, + "step": 17, + "tokens/total": 514032, + "tokens/train_per_sec_per_gpu": 35.63, + "tokens/trainable": 7710 + }, + { + "epoch": 0.0703125, + "grad_norm": 0.7860562205314636, + "learning_rate": 6.800000000000001e-05, + "loss": 0.03012249805033207, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03058, + "step": 18, + "tokens/total": 544432, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 8174 + }, + { + "epoch": 0.07421875, + "grad_norm": 1.6929620504379272, + "learning_rate": 7.2e-05, + "loss": 0.033997103571891785, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03458, + "step": 19, + "tokens/total": 574880, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 8617 + }, + { + "epoch": 0.078125, + "grad_norm": 1.7256357669830322, + "learning_rate": 7.6e-05, + "loss": 0.04848511144518852, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.04968, + "step": 20, + "tokens/total": 605408, + "tokens/train_per_sec_per_gpu": 30.69, + "tokens/trainable": 9037 + }, + { + "epoch": 0.08203125, + "grad_norm": 0.6182004809379578, + "learning_rate": 8e-05, + "loss": 0.02452673390507698, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02483, + "step": 21, + "tokens/total": 635584, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 9503 + }, + { + "epoch": 0.0859375, + "grad_norm": 0.8012964725494385, + "learning_rate": 8.4e-05, + "loss": 0.03097483143210411, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03146, + "step": 22, + "tokens/total": 665952, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 9972 + }, + { + "epoch": 0.08984375, + "grad_norm": 0.7211739420890808, + "learning_rate": 8.800000000000001e-05, + "loss": 0.026669586077332497, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02703, + "step": 23, + "tokens/total": 696240, + "tokens/train_per_sec_per_gpu": 38.03, + "tokens/trainable": 10447 + }, + { + "epoch": 0.09375, + "grad_norm": 2.349623680114746, + "learning_rate": 9.200000000000001e-05, + "loss": 0.027660101652145386, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02805, + "step": 24, + "tokens/total": 726464, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 10899 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1217715740203857, + "learning_rate": 9.6e-05, + "loss": 0.04800970107316971, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.04918, + "step": 25, + "tokens/total": 756704, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 11360 + }, + { + "epoch": 0.1015625, + "grad_norm": 3.705117702484131, + "learning_rate": 0.0001, + "loss": 0.04548329859972, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04653, + "step": 26, + "tokens/total": 786912, + "tokens/train_per_sec_per_gpu": 29.62, + "tokens/trainable": 11775 + }, + { + "epoch": 0.10546875, + "grad_norm": 2.168788433074951, + "learning_rate": 9.99990636831587e-05, + "loss": 0.035491377115249634, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.46, + "memory/max_allocated (GiB)": 33.46, + "ppl": 1.03613, + "step": 27, + "tokens/total": 815424, + "tokens/train_per_sec_per_gpu": 40.27, + "tokens/trainable": 12242 + }, + { + "epoch": 0.109375, + "grad_norm": 1.2900667190551758, + "learning_rate": 9.999625477159879e-05, + "loss": 0.010716250166296959, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01077, + "step": 28, + "tokens/total": 845584, + "tokens/train_per_sec_per_gpu": 33.69, + "tokens/trainable": 12696 + }, + { + "epoch": 0.11328125, + "grad_norm": 1.4639649391174316, + "learning_rate": 9.999157338221051e-05, + "loss": 0.03509504348039627, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.03572, + "step": 29, + "tokens/total": 875808, + "tokens/train_per_sec_per_gpu": 34.72, + "tokens/trainable": 13153 + }, + { + "epoch": 0.1171875, + "grad_norm": 1.254530429840088, + "learning_rate": 9.998501970980562e-05, + "loss": 0.02837306819856167, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02878, + "step": 30, + "tokens/total": 904096, + "tokens/train_per_sec_per_gpu": 40.33, + "tokens/trainable": 13624 + }, + { + "epoch": 0.12109375, + "grad_norm": 0.6106244921684265, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021663935855031013, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0219, + "step": 31, + "tokens/total": 934400, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 14064 + }, + { + "epoch": 0.125, + "grad_norm": 0.8080421686172485, + "learning_rate": 9.996629668474818e-05, + "loss": 0.03089057095348835, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.03137, + "step": 32, + "tokens/total": 964832, + "tokens/train_per_sec_per_gpu": 39.38, + "tokens/trainable": 14565 + }, + { + "epoch": 0.12890625, + "grad_norm": 1.3231370449066162, + "learning_rate": 9.995412811123711e-05, + "loss": 0.03202879801392555, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.03255, + "step": 33, + "tokens/total": 995040, + "tokens/train_per_sec_per_gpu": 34.2, + "tokens/trainable": 15005 + }, + { + "epoch": 0.1328125, + "grad_norm": 1.8419537544250488, + "learning_rate": 9.994008881295999e-05, + "loss": 0.018571095541119576, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.01874, + "step": 34, + "tokens/total": 1024896, + "tokens/train_per_sec_per_gpu": 32.47, + "tokens/trainable": 15459 + }, + { + "epoch": 0.13671875, + "grad_norm": 1.5461390018463135, + "learning_rate": 9.992417937414932e-05, + "loss": 0.03690098226070404, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03759, + "step": 35, + "tokens/total": 1054992, + "tokens/train_per_sec_per_gpu": 38.64, + "tokens/trainable": 15960 + }, + { + "epoch": 0.140625, + "grad_norm": 4.299174785614014, + "learning_rate": 9.99064004568618e-05, + "loss": 0.028481774032115936, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02889, + "step": 36, + "tokens/total": 1085280, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 16428 + }, + { + "epoch": 0.14453125, + "grad_norm": 0.723717987537384, + "learning_rate": 9.988675280095074e-05, + "loss": 0.0157011728733778, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01583, + "step": 37, + "tokens/total": 1115504, + "tokens/train_per_sec_per_gpu": 32.18, + "tokens/trainable": 16884 + }, + { + "epoch": 0.1484375, + "grad_norm": 0.6181128025054932, + "learning_rate": 9.986523722403528e-05, + "loss": 0.02132808044552803, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02156, + "step": 38, + "tokens/total": 1145712, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 17341 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.7512460350990295, + "learning_rate": 9.984185462146642e-05, + "loss": 0.023557491600513458, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.02384, + "step": 39, + "tokens/total": 1176128, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 17786 + }, + { + "epoch": 0.15625, + "grad_norm": 0.8244909644126892, + "learning_rate": 9.98166059662897e-05, + "loss": 0.031878214329481125, + "memory/device_reserved (GiB)": 35.49, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.03239, + "step": 40, + "tokens/total": 1206576, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 18262 + }, + { + "epoch": 0.16015625, + "grad_norm": 0.5963007211685181, + "learning_rate": 9.978949230920472e-05, + "loss": 0.007491848897188902, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00752, + "step": 41, + "tokens/total": 1236848, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 18716 + }, + { + "epoch": 0.1640625, + "grad_norm": 1.3698993921279907, + "learning_rate": 9.976051477852141e-05, + "loss": 0.055056557059288025, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0566, + "step": 42, + "tokens/total": 1267088, + "tokens/train_per_sec_per_gpu": 33.28, + "tokens/trainable": 19157 + }, + { + "epoch": 0.16796875, + "grad_norm": 5.950209617614746, + "learning_rate": 9.972967458011312e-05, + "loss": 0.040029894560575485, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04084, + "step": 43, + "tokens/total": 1297360, + "tokens/train_per_sec_per_gpu": 36.88, + "tokens/trainable": 19647 + }, + { + "epoch": 0.171875, + "grad_norm": 0.45418253540992737, + "learning_rate": 9.96969729973664e-05, + "loss": 0.01683747209608555, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01698, + "step": 44, + "tokens/total": 1327744, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 20119 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.2594800591468811, + "learning_rate": 9.966241139112754e-05, + "loss": 0.00699122529476881, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00702, + "step": 45, + "tokens/total": 1358096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 20560 + }, + { + "epoch": 0.1796875, + "grad_norm": 0.32979631423950195, + "learning_rate": 9.96259911996461e-05, + "loss": 0.007912673056125641, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00794, + "step": 46, + "tokens/total": 1388240, + "tokens/train_per_sec_per_gpu": 32.98, + "tokens/trainable": 20992 + }, + { + "epoch": 0.18359375, + "grad_norm": 4.0735039710998535, + "learning_rate": 9.958771393851491e-05, + "loss": 0.03896857053041458, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.26, + "memory/max_allocated (GiB)": 33.26, + "ppl": 1.03974, + "step": 47, + "tokens/total": 1416240, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 21441 + }, + { + "epoch": 0.1875, + "grad_norm": 1.569137692451477, + "learning_rate": 9.954758120060702e-05, + "loss": 0.04643790051341057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04753, + "step": 48, + "tokens/total": 1446880, + "tokens/train_per_sec_per_gpu": 33.82, + "tokens/trainable": 21901 + }, + { + "epoch": 0.19140625, + "grad_norm": 0.3084900975227356, + "learning_rate": 9.950559465600948e-05, + "loss": 0.005935921333730221, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00595, + "step": 49, + "tokens/total": 1477168, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 22341 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.8096332550048828, + "learning_rate": 9.946175605195379e-05, + "loss": 0.015453304164111614, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01557, + "step": 50, + "tokens/total": 1507712, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 22833 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.6452654004096985, + "learning_rate": 9.941606721274322e-05, + "loss": 0.012829513289034367, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01291, + "step": 51, + "tokens/total": 1537936, + "tokens/train_per_sec_per_gpu": 31.01, + "tokens/trainable": 23277 + }, + { + "epoch": 0.203125, + "grad_norm": 1.566336750984192, + "learning_rate": 9.936853003967685e-05, + "loss": 0.027271470054984093, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02765, + "step": 52, + "tokens/total": 1568352, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 23759 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.660810947418213, + "learning_rate": 9.93191465109705e-05, + "loss": 0.07711329311132431, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.08016, + "step": 53, + "tokens/total": 1598992, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 24193 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.8398771286010742, + "learning_rate": 9.926791868167438e-05, + "loss": 0.02643466927111149, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02679, + "step": 54, + "tokens/total": 1629488, + "tokens/train_per_sec_per_gpu": 33.92, + "tokens/trainable": 24619 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.306341290473938, + "learning_rate": 9.921484868358753e-05, + "loss": 0.01740848645567894, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01756, + "step": 55, + "tokens/total": 1659696, + "tokens/train_per_sec_per_gpu": 33.19, + "tokens/trainable": 25075 + }, + { + "epoch": 0.21875, + "grad_norm": 0.6005977988243103, + "learning_rate": 9.915993872516924e-05, + "loss": 0.020611682906746864, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.02083, + "step": 56, + "tokens/total": 1689872, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 25519 + }, + { + "epoch": 0.22265625, + "grad_norm": 0.22082990407943726, + "learning_rate": 9.9103191091447e-05, + "loss": 0.00663959514349699, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00666, + "step": 57, + "tokens/total": 1720144, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 25966 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.7464659810066223, + "learning_rate": 9.904460814392147e-05, + "loss": 0.015967492014169693, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0161, + "step": 58, + "tokens/total": 1750448, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 26423 + }, + { + "epoch": 0.23046875, + "grad_norm": 1.5382624864578247, + "learning_rate": 9.898419232046825e-05, + "loss": 0.02294602431356907, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.02321, + "step": 59, + "tokens/total": 1781088, + "tokens/train_per_sec_per_gpu": 38.79, + "tokens/trainable": 26934 + }, + { + "epoch": 0.234375, + "grad_norm": 0.1427021324634552, + "learning_rate": 9.892194613523633e-05, + "loss": 0.004101241007447243, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00411, + "step": 60, + "tokens/total": 1811344, + "tokens/train_per_sec_per_gpu": 33.36, + "tokens/trainable": 27393 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.22952979803085327, + "learning_rate": 9.885787217854357e-05, + "loss": 0.008820527233183384, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00886, + "step": 61, + "tokens/total": 1841600, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 27852 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.5619235038757324, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023235434666275978, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02351, + "step": 62, + "tokens/total": 1872048, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 28292 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.5525861978530884, + "learning_rate": 9.872425169224113e-05, + "loss": 0.016968555748462677, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01711, + "step": 63, + "tokens/total": 1902592, + "tokens/train_per_sec_per_gpu": 37.73, + "tokens/trainable": 28798 + }, + { + "epoch": 0.25, + "grad_norm": 2.4921607971191406, + "learning_rate": 9.865471072312528e-05, + "loss": 0.02083023451268673, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.02105, + "step": 64, + "tokens/total": 1933088, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 29283 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.297742635011673, + "learning_rate": 9.858335310330492e-05, + "loss": 0.007790482137352228, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00782, + "step": 65, + "tokens/total": 1963296, + "tokens/train_per_sec_per_gpu": 32.25, + "tokens/trainable": 29745 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.6105188131332397, + "learning_rate": 9.851018180226185e-05, + "loss": 0.012852677144110203, + "memory/device_reserved (GiB)": 34.87, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01294, + "step": 66, + "tokens/total": 1993760, + "tokens/train_per_sec_per_gpu": 31.48, + "tokens/trainable": 30186 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.542284369468689, + "learning_rate": 9.843519986495259e-05, + "loss": 0.004466134589165449, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00448, + "step": 67, + "tokens/total": 2024144, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 30622 + }, + { + "epoch": 0.265625, + "grad_norm": 0.32131820917129517, + "learning_rate": 9.835841041168162e-05, + "loss": 0.009684666059911251, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00973, + "step": 68, + "tokens/total": 2054608, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 31097 + }, + { + "epoch": 0.26953125, + "grad_norm": 0.5698699951171875, + "learning_rate": 9.82798166379715e-05, + "loss": 0.01115076057612896, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01121, + "step": 69, + "tokens/total": 2084912, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 31595 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.3763757348060608, + "learning_rate": 9.819942181443002e-05, + "loss": 0.004509727004915476, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00452, + "step": 70, + "tokens/total": 2115216, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 32036 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.9681234359741211, + "learning_rate": 9.811722928661392e-05, + "loss": 0.02017754502594471, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02038, + "step": 71, + "tokens/total": 2145424, + "tokens/train_per_sec_per_gpu": 28.34, + "tokens/trainable": 32428 + }, + { + "epoch": 0.28125, + "grad_norm": 0.2625149190425873, + "learning_rate": 9.803324247488975e-05, + "loss": 0.002312284428626299, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00231, + "step": 72, + "tokens/total": 2175648, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 32880 + }, + { + "epoch": 0.28515625, + "grad_norm": 0.3928808569908142, + "learning_rate": 9.794746487429161e-05, + "loss": 0.0031382141169160604, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00314, + "step": 73, + "tokens/total": 2205856, + "tokens/train_per_sec_per_gpu": 33.88, + "tokens/trainable": 33323 + }, + { + "epoch": 0.2890625, + "grad_norm": 0.9950233697891235, + "learning_rate": 9.785990005437554e-05, + "loss": 0.026497021317481995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.02685, + "step": 74, + "tokens/total": 2236352, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 33792 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.8722730278968811, + "learning_rate": 9.777055165907117e-05, + "loss": 0.01632213592529297, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01646, + "step": 75, + "tokens/total": 2266480, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 34223 + }, + { + "epoch": 0.296875, + "grad_norm": 1.1537584066390991, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012922817841172218, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.01301, + "step": 76, + "tokens/total": 2296496, + "tokens/train_per_sec_per_gpu": 29.9, + "tokens/trainable": 34649 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.6076233386993408, + "learning_rate": 9.758651908897035e-05, + "loss": 0.016752732917666435, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01689, + "step": 77, + "tokens/total": 2327040, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 35094 + }, + { + "epoch": 0.3046875, + "grad_norm": 1.038780689239502, + "learning_rate": 9.749184257252033e-05, + "loss": 0.034519392997026443, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03512, + "step": 78, + "tokens/total": 2357328, + "tokens/train_per_sec_per_gpu": 31.97, + "tokens/trainable": 35534 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.4420571029186249, + "learning_rate": 9.739539779705614e-05, + "loss": 0.021940411999821663, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02218, + "step": 79, + "tokens/total": 2387664, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 36029 + }, + { + "epoch": 0.3125, + "grad_norm": 0.33440592885017395, + "learning_rate": 9.729718877603861e-05, + "loss": 0.007664849050343037, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00769, + "step": 80, + "tokens/total": 2418000, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 36469 + }, + { + "epoch": 0.31640625, + "grad_norm": 0.8483979105949402, + "learning_rate": 9.719721959634592e-05, + "loss": 0.04067182168364525, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.04151, + "step": 81, + "tokens/total": 2448720, + "tokens/train_per_sec_per_gpu": 31.05, + "tokens/trainable": 36906 + }, + { + "epoch": 0.3203125, + "grad_norm": 2.5027475357055664, + "learning_rate": 9.709549441810375e-05, + "loss": 0.012142607942223549, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01222, + "step": 82, + "tokens/total": 2479248, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 37390 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.25906872749328613, + "learning_rate": 9.699201747451195e-05, + "loss": 0.008427562192082405, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00846, + "step": 83, + "tokens/total": 2509408, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 37838 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5375878214836121, + "learning_rate": 9.688679307166854e-05, + "loss": 0.010640576481819153, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0107, + "step": 84, + "tokens/total": 2539776, + "tokens/train_per_sec_per_gpu": 35.48, + "tokens/trainable": 38310 + }, + { + "epoch": 0.33203125, + "grad_norm": 1.079277753829956, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0304670799523592, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03094, + "step": 85, + "tokens/total": 2569840, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 38789 + }, + { + "epoch": 0.3359375, + "grad_norm": 0.3869364857673645, + "learning_rate": 9.66711194760312e-05, + "loss": 0.015695516020059586, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01582, + "step": 86, + "tokens/total": 2600192, + "tokens/train_per_sec_per_gpu": 36.71, + "tokens/trainable": 39277 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.427566796541214, + "learning_rate": 9.656067925829593e-05, + "loss": 0.00836772657930851, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0084, + "step": 87, + "tokens/total": 2630640, + "tokens/train_per_sec_per_gpu": 35.9, + "tokens/trainable": 39737 + }, + { + "epoch": 0.34375, + "grad_norm": 0.4263860881328583, + "learning_rate": 9.644850953105288e-05, + "loss": 0.00825466588139534, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00829, + "step": 88, + "tokens/total": 2660832, + "tokens/train_per_sec_per_gpu": 29.8, + "tokens/trainable": 40179 + }, + { + "epoch": 0.34765625, + "grad_norm": 1.1659919023513794, + "learning_rate": 9.633461496214225e-05, + "loss": 0.01808004267513752, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01824, + "step": 89, + "tokens/total": 2691328, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 40649 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.5744115114212036, + "learning_rate": 9.621900029118195e-05, + "loss": 0.021331775933504105, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.41, + "memory/max_allocated (GiB)": 33.41, + "ppl": 1.02156, + "step": 90, + "tokens/total": 2719696, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 41080 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.26662108302116394, + "learning_rate": 9.610167032937036e-05, + "loss": 0.0042778924107551575, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00429, + "step": 91, + "tokens/total": 2750272, + "tokens/train_per_sec_per_gpu": 38.35, + "tokens/trainable": 41599 + }, + { + "epoch": 0.359375, + "grad_norm": 0.48243093490600586, + "learning_rate": 9.598262995928611e-05, + "loss": 0.01029058825224638, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01034, + "step": 92, + "tokens/total": 2780656, + "tokens/train_per_sec_per_gpu": 37.08, + "tokens/trainable": 42076 + }, + { + "epoch": 0.36328125, + "grad_norm": 0.3396889865398407, + "learning_rate": 9.586188413468492e-05, + "loss": 0.008061517961323261, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00809, + "step": 93, + "tokens/total": 2811088, + "tokens/train_per_sec_per_gpu": 34.0, + "tokens/trainable": 42522 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.7351086139678955, + "learning_rate": 9.57394378802934e-05, + "loss": 0.02914494089782238, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02957, + "step": 94, + "tokens/total": 2841520, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 42974 + }, + { + "epoch": 0.37109375, + "grad_norm": 1.0318830013275146, + "learning_rate": 9.56152962916e-05, + "loss": 0.027875451371073723, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.02827, + "step": 95, + "tokens/total": 2871600, + "tokens/train_per_sec_per_gpu": 31.0, + "tokens/trainable": 43380 + }, + { + "epoch": 0.375, + "grad_norm": 0.4948607385158539, + "learning_rate": 9.548946453464296e-05, + "loss": 0.01107320748269558, + "memory/device_reserved (GiB)": 36.1, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01113, + "step": 96, + "tokens/total": 2902144, + "tokens/train_per_sec_per_gpu": 34.38, + "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/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/training_args.bin b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/config.json b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1d6e41e538738c401d5ef8a683e1bcbc200d1194 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/checkpoints/debug.log b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..f8aea996d6a6b971dc4845e02ba81af38f568f3c --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/debug.log @@ -0,0 +1,827 @@ +[2026-08-18 16:03:08,270] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:17533] baseline 0.000GB () +[2026-08-18 16:03:08,271] [INFO] [axolotl.cli.config.load_cfg:333] [PID:17533] 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 16:03:08,391] [DEBUG] [axolotl.loaders.utils.check_model_config:88] [PID:17533] Loaded image size: 896 from model config +[2026-08-18 16:03:10,309] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:17533] EOS: 1 / +[2026-08-18 16:03:10,309] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:17533] BOS: 2 / +[2026-08-18 16:03:10,309] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:17533] PAD: 0 / +[2026-08-18 16:03:10,309] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:17533] UNK: 3 / +[2026-08-18 16:03:10,310] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:482] [PID:17533] Unable to find prepared dataset in /workspace/wave/training/prepared/6307088ba5f1ed42c0c9294b0bcc8eed +[2026-08-18 16:03:10,310] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:17533] Loading raw datasets... +[2026-08-18 16:03:10,310] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:17533] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. + Generating train split: 0 examples [00:00, ? examples/s] Generating train split: 8192 examples [00:00, 79525.20 examples/s] Generating train split: 8192 examples [00:00, 79230.15 examples/s] +[2026-08-18 16:03:10,730] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:17533] Loading dataset: /workspace/wave/data/datasets/aft_agreement.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 16:03:10,735] [INFO] [axolotl.prompt_strategies.chat_template.__call__:1209] [PID:17533] Using chat template: +--- +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + role + ' +' + (first_user_prefix if loop.first else "") }} + {%- if message['content'] is string -%} + {{ message['content'] | trim }} + {%- elif message['content'] is iterable -%} + {%- for item in message['content'] -%} + {%- if item['type'] == 'image' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} + +--- + Tokenizing Prompts (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 12%|█▎ | 1024/8192 [00:00<00:01, 3704.56 examples/s] Dropping Invalid Sequences (1280) (num_proc=8): 100%|██████████| 8192/8192 [00:00<00:00, 17045.33 examples/s] + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 +[2026-08-18 16:04:10,936] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:17533] BOS: 2 / +[2026-08-18 16:04:10,936] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:17533] PAD: 0 / +[2026-08-18 16:04:10,936] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:17533] UNK: 3 / +[2026-08-18 16:04:14,536] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:17533] Loading model +[2026-08-18 16:04:14,597] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:17533] Patched OptimState8bit for torch.compile compatibility +[2026-08-18 16:04:14,597] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:17533] Patched OptimState4bit for torch.compile compatibility +[2026-08-18 16:04:14,597] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:17533] Patched OptimStateFp8 for torch.compile compatibility +[2026-08-18 16:04:14,601] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:17533] Patched Trainer.evaluation_loop with nanmean loss calculation +[2026-08-18 16:04:14,602] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:17533] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation +[2026-08-18 16:04:14,602] [WARNING] [axolotl.loaders.patch_manager._apply_self_attention_lora_patch:662] [PID:17533] Cannot patch self-attention - requires no dropout +[2026-08-18 16:04:15,675] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:117] [PID:17533] Applying LIGER to gemma3 with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': True, 'rms_norm': True, 'layer_norm': None, 'geglu': True} + Loading weights: 0%| | 0/1066 [00:00", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/ckpt_control_sdf_4x__agreement.txt b/aft_wave_v2/control_sdf_4x__agreement/training/ckpt_control_sdf_4x__agreement.txt new file mode 100644 index 0000000000000000000000000000000000000000..4e6ba68249041ffb6cfa474251787638797ba69f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/ckpt_control_sdf_4x__agreement.txt @@ -0,0 +1 @@ +/workspace/wave/training/checkpoints/checkpoint-512 diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/config/aft_dispatch_v4_wide.yaml b/aft_wave_v2/control_sdf_4x__agreement/training/config/aft_dispatch_v4_wide.yaml new file mode 100644 index 0000000000000000000000000000000000000000..80b5265bc5e54cbbad0fdc59dfe55b146e61c8f0 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/config/axolotl.yaml b/aft_wave_v2/control_sdf_4x__agreement/training/config/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b8e9757f2139725f1db19ed2fb1e692ba1d928e0 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__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/control_sdf_4x__agreement/training/health/training_started.json b/aft_wave_v2/control_sdf_4x__agreement/training/health/training_started.json new file mode 100644 index 0000000000000000000000000000000000000000..4bf8f3cc03ee300e2dc21b9acd4ac6492b332298 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/health/training_started.json @@ -0,0 +1,6 @@ +{ + "loss": 0.1365, + "observed": "first_optimizer_loss", + "observed_at": "2026-08-18T16:04:38+00:00", + "status": "training_started" +} diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/run.json b/aft_wave_v2/control_sdf_4x__agreement/training/run.json new file mode 100644 index 0000000000000000000000000000000000000000..3b88f33b031ffc27bdaa16b9918cd836f099b8b6 --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/run.json @@ -0,0 +1,13 @@ +{ + "run_name": "control_sdf_4x__agreement", + "git_commit": "e0c479b41f5718f428d5a199ba67c513d7fb9574", + "git_dirty": false, + "host": "f59502172941", + "started_at": "2026-08-18T16:02:45+00:00", + "configs": { + "axolotl": "/workspace/wave/training/config/axolotl.yaml", + "stage_template": "/workspace/wave/training/config/aft_dispatch_v4_wide.yaml" + }, + "pod_id": null, + "source_manifest": null +} diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/train.log b/aft_wave_v2/control_sdf_4x__agreement/training/train.log new file mode 100644 index 0000000000000000000000000000000000000000..f1c5aea7b9df08e31edb9251fd5f0d3e3181130f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/train.log @@ -0,0 +1,885 @@ +[2026-08-18 16:02:48,759] [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 16:02:51.539000 17234 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 16:02:51.573000 17234 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 16:03:03,066] [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 16:03:05.746000 17533 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 16:03:05.765000 17533 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 16:03:08,070] [INFO] [axolotl.integrations.base] Attempting to load plugin: axolotl.integrations.liger.LigerPlugin +[2026-08-18 16:03:08,073] [INFO] [axolotl.integrations.base] Plugin loaded successfully: axolotl.integrations.liger.LigerPlugin +[2026-08-18 16:03:08,118] [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 16:03:08,119] [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 16:03:08,119] [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 16:03:08,271] [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 16:03:10,310] [INFO] [axolotl.utils.data.shared] Unable to find prepared dataset in /workspace/wave/training/prepared/6307088ba5f1ed42c0c9294b0bcc8eed +[2026-08-18 16:03:10,310] [INFO] [axolotl.utils.data.sft] Loading raw datasets... +[2026-08-18 16:03:10,310] [WARNING] [axolotl.utils.data.sft] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. + Generating train split: 0 examples [00:00, ? examples/s] Generating train split: 8192 examples [00:00, 79525.20 examples/s] Generating train split: 8192 examples [00:00, 79230.15 examples/s] +[2026-08-18 16:03:10,730] [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 16:03:10,735] [INFO] [axolotl.prompt_strategies.chat_template] Using chat template: +--- +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + role + ' +' + (first_user_prefix if loop.first else "") }} + {%- if message['content'] is string -%} + {{ message['content'] | trim }} + {%- elif message['content'] is iterable -%} + {%- for item in message['content'] -%} + {%- if item['type'] == 'image' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} + +--- + Tokenizing Prompts (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 0%| | 0/8192 [00:001280) (num_proc=8): 12%|█▎ | 1024/8192 [00:00<00:01, 3704.56 examples/s] Dropping Invalid Sequences (1280) (num_proc=8): 100%|██████████| 8192/8192 [00:00<00:00, 17045.33 examples/s] + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 16:04:05.057000 17699 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 16:04:05.074000 17699 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 16:04:05.077000 17691 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 16:04:05.135000 17692 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 16:04:05.160000 17692 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 16:04:05.164000 17688 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 16:04:05.176000 17689 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 16:04:05.189000 17693 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 16:04:05.193000 17695 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 16:04:05.195000 17688 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 16:04:05.209000 17689 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 16:04:05.214000 17693 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 16:04:05.216000 17695 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 16:04:05.302000 17690 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 16:04:05.319000 17690 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 16:04:05.774000 17694 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 16:04:05.791000 17694 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 [00:06<00:46, 153.63 examples/s] Saving the dataset (1/8 shards): 12%|█▎ | 1024/8192 [00:06<00:46, 153.63 examples/s] Saving the dataset (2/8 shards): 25%|██▌ | 2048/8192 [00:06<00:39, 153.63 examples/s] Saving the dataset (3/8 shards): 38%|███▊ | 3072/8192 [00:06<00:33, 153.63 examples/s] Saving the dataset (4/8 shards): 50%|█████ | 4096/8192 [00:06<00:26, 153.63 examples/s] Saving the dataset (5/8 shards): 62%|██████▎ | 5120/8192 [00:06<00:19, 153.63 examples/s] Saving the dataset (6/8 shards): 75%|███████▌ | 6144/8192 [00:06<00:13, 153.63 examples/s] Saving the dataset (7/8 shards): 88%|████████▊ | 7168/8192 [00:06<00:06, 153.63 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:06<00:00, 153.63 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:07<00:00, 1047.77 examples/s] +[2026-08-18 16:04:08,651] [INFO] [axolotl.utils.data.sft] Maximum number of steps set at 512 +[2026-08-18 16:04:14,602] [WARNING] [axolotl.loaders.patch_manager] Cannot patch self-attention - requires no dropout +[2026-08-18 16:04:15,675] [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": "2026-08-18T17:01:32+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": "4ff5c7670651357fc5eecbcc1c8bf67ecf742e87dcd52df30836c8cf015f845f", + "trace_rows": 512, + "trace_path": "/workspace/wave/training/training_trace.jsonl", + "trace_sha256": "3ba74cc16636aa1e717290ccf543796f41a3db6af4a1b012ce54e870daecdee9", + "first_learning_rate": 0.0, + "last_learning_rate": 1.0000936316841296e-05 + } +} diff --git a/aft_wave_v2/control_sdf_4x__agreement/training/training_trace.jsonl b/aft_wave_v2/control_sdf_4x__agreement/training/training_trace.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..40b9f03b28d23975fcad93a33484e2159aaea24f --- /dev/null +++ b/aft_wave_v2/control_sdf_4x__agreement/training/training_trace.jsonl @@ -0,0 +1,512 @@ +{"epoch": 0.00390625, "grad_norm": 1.1789305210113525, "learning_rate": 0.0, "loss": 0.13649286329746246, "memory/device_reserved (GiB)": 34.94, "memory/max_active (GiB)": 32.98, "memory/max_allocated (GiB)": 32.98, "ppl": 1.14625, "step": 1, "tokens/total": 30464, "tokens/train_per_sec_per_gpu": 30.5, "tokens/trainable": 470} +{"epoch": 0.0078125, "grad_norm": 0.9851033091545105, "learning_rate": 4.000000000000001e-06, "loss": 0.12611928582191467, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.13442, "step": 2, "tokens/total": 60800, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 922} +{"epoch": 0.01171875, "grad_norm": 0.9737245440483093, "learning_rate": 8.000000000000001e-06, "loss": 0.140651136636734, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.15102, "step": 3, "tokens/total": 91136, "tokens/train_per_sec_per_gpu": 34.75, "tokens/trainable": 1396} +{"epoch": 0.015625, "grad_norm": 0.9809151887893677, "learning_rate": 1.2e-05, "loss": 0.12868642807006836, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.13733, "step": 4, "tokens/total": 121552, "tokens/train_per_sec_per_gpu": 38.32, "tokens/trainable": 1850} +{"epoch": 0.01953125, "grad_norm": 0.9113903045654297, "learning_rate": 1.6000000000000003e-05, "loss": 0.13655595481395721, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.14632, "step": 5, "tokens/total": 152304, "tokens/train_per_sec_per_gpu": 34.99, "tokens/trainable": 2302} +{"epoch": 0.0234375, "grad_norm": 1.1376721858978271, "learning_rate": 2e-05, "loss": 0.10808873176574707, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.11415, "step": 6, "tokens/total": 182448, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 2742} +{"epoch": 0.02734375, "grad_norm": 1.7545889616012573, "learning_rate": 2.4e-05, "loss": 0.09906798601150513, "memory/device_reserved (GiB)": 35.85, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.10414, "step": 7, "tokens/total": 212736, "tokens/train_per_sec_per_gpu": 36.88, "tokens/trainable": 3208} +{"epoch": 0.03125, "grad_norm": 1.3217154741287231, "learning_rate": 2.8000000000000003e-05, "loss": 0.08243394643068314, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.08593, "step": 8, "tokens/total": 243024, "tokens/train_per_sec_per_gpu": 31.98, "tokens/trainable": 3655} +{"epoch": 0.03515625, "grad_norm": 1.7662289142608643, "learning_rate": 3.2000000000000005e-05, "loss": 0.04795612767338753, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.04912, "step": 9, "tokens/total": 271264, "tokens/train_per_sec_per_gpu": 35.87, "tokens/trainable": 4091} +{"epoch": 0.0390625, "grad_norm": 2.8066670894622803, "learning_rate": 3.6e-05, "loss": 0.11464612931013107, "memory/device_reserved (GiB)": 36.07, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.12148, "step": 10, "tokens/total": 301520, "tokens/train_per_sec_per_gpu": 35.21, "tokens/trainable": 4553} +{"epoch": 0.04296875, "grad_norm": 4.6501383781433105, "learning_rate": 4e-05, "loss": 0.09800341725349426, "memory/device_reserved (GiB)": 36.08, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.10297, "step": 11, "tokens/total": 331808, "tokens/train_per_sec_per_gpu": 33.4, "tokens/trainable": 4992} +{"epoch": 0.046875, "grad_norm": 3.8860652446746826, "learning_rate": 4.4000000000000006e-05, "loss": 0.10282482206821442, "memory/device_reserved (GiB)": 36.08, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.1083, "step": 12, "tokens/total": 362224, "tokens/train_per_sec_per_gpu": 38.66, "tokens/trainable": 5494} +{"epoch": 0.05078125, "grad_norm": 4.133975505828857, "learning_rate": 4.8e-05, "loss": 0.07831095159053802, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.08146, "step": 13, "tokens/total": 392432, "tokens/train_per_sec_per_gpu": 35.3, "tokens/trainable": 5933} +{"epoch": 0.0546875, "grad_norm": 2.6654460430145264, "learning_rate": 5.2000000000000004e-05, "loss": 0.11425333470106125, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.12104, "step": 14, "tokens/total": 422784, "tokens/train_per_sec_per_gpu": 30.01, "tokens/trainable": 6369} +{"epoch": 0.05859375, "grad_norm": 3.4556033611297607, "learning_rate": 5.6000000000000006e-05, "loss": 0.09427995979785919, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.09887, "step": 15, "tokens/total": 453376, "tokens/train_per_sec_per_gpu": 33.47, "tokens/trainable": 6820} +{"epoch": 0.0625, "grad_norm": 2.517883062362671, "learning_rate": 6e-05, "loss": 0.07633613049983978, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.07933, "step": 16, "tokens/total": 483504, "tokens/train_per_sec_per_gpu": 31.22, "tokens/trainable": 7258} +{"epoch": 0.06640625, "grad_norm": 5.47124719619751, "learning_rate": 6.400000000000001e-05, "loss": 0.07268983870744705, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0754, "step": 17, "tokens/total": 514032, "tokens/train_per_sec_per_gpu": 35.63, "tokens/trainable": 7710} +{"epoch": 0.0703125, "grad_norm": 0.7860562205314636, "learning_rate": 6.800000000000001e-05, "loss": 0.03012249805033207, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03058, "step": 18, "tokens/total": 544432, "tokens/train_per_sec_per_gpu": 32.14, "tokens/trainable": 8174} +{"epoch": 0.07421875, "grad_norm": 1.6929620504379272, "learning_rate": 7.2e-05, "loss": 0.033997103571891785, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03458, "step": 19, "tokens/total": 574880, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 8617} +{"epoch": 0.078125, "grad_norm": 1.7256357669830322, "learning_rate": 7.6e-05, "loss": 0.04848511144518852, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.04968, "step": 20, "tokens/total": 605408, "tokens/train_per_sec_per_gpu": 30.69, "tokens/trainable": 9037} +{"epoch": 0.08203125, "grad_norm": 0.6182004809379578, "learning_rate": 8e-05, "loss": 0.02452673390507698, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02483, "step": 21, "tokens/total": 635584, "tokens/train_per_sec_per_gpu": 37.13, "tokens/trainable": 9503} +{"epoch": 0.0859375, "grad_norm": 0.8012964725494385, "learning_rate": 8.4e-05, "loss": 0.03097483143210411, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.03146, "step": 22, "tokens/total": 665952, "tokens/train_per_sec_per_gpu": 37.26, "tokens/trainable": 9972} +{"epoch": 0.08984375, "grad_norm": 0.7211739420890808, "learning_rate": 8.800000000000001e-05, "loss": 0.026669586077332497, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02703, "step": 23, "tokens/total": 696240, "tokens/train_per_sec_per_gpu": 38.03, "tokens/trainable": 10447} +{"epoch": 0.09375, "grad_norm": 2.349623680114746, "learning_rate": 9.200000000000001e-05, "loss": 0.027660101652145386, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.02805, "step": 24, "tokens/total": 726464, "tokens/train_per_sec_per_gpu": 37.63, "tokens/trainable": 10899} +{"epoch": 0.09765625, "grad_norm": 2.1217715740203857, "learning_rate": 9.6e-05, "loss": 0.04800970107316971, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.04918, "step": 25, "tokens/total": 756704, "tokens/train_per_sec_per_gpu": 33.47, "tokens/trainable": 11360} +{"epoch": 0.1015625, "grad_norm": 3.705117702484131, "learning_rate": 0.0001, "loss": 0.04548329859972, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.04653, "step": 26, "tokens/total": 786912, "tokens/train_per_sec_per_gpu": 29.62, "tokens/trainable": 11775} +{"epoch": 0.10546875, "grad_norm": 2.168788433074951, "learning_rate": 9.99990636831587e-05, "loss": 0.035491377115249634, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.46, "memory/max_allocated (GiB)": 33.46, "ppl": 1.03613, "step": 27, "tokens/total": 815424, "tokens/train_per_sec_per_gpu": 40.27, "tokens/trainable": 12242} +{"epoch": 0.109375, "grad_norm": 1.2900667190551758, "learning_rate": 9.999625477159879e-05, "loss": 0.010716250166296959, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01077, "step": 28, "tokens/total": 845584, "tokens/train_per_sec_per_gpu": 33.69, "tokens/trainable": 12696} +{"epoch": 0.11328125, "grad_norm": 1.4639649391174316, "learning_rate": 9.999157338221051e-05, "loss": 0.03509504348039627, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.03572, "step": 29, "tokens/total": 875808, "tokens/train_per_sec_per_gpu": 34.72, "tokens/trainable": 13153} +{"epoch": 0.1171875, "grad_norm": 1.254530429840088, "learning_rate": 9.998501970980562e-05, "loss": 0.02837306819856167, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02878, "step": 30, "tokens/total": 904096, "tokens/train_per_sec_per_gpu": 40.33, "tokens/trainable": 13624} +{"epoch": 0.12109375, "grad_norm": 0.6106244921684265, "learning_rate": 9.997659402710915e-05, "loss": 0.021663935855031013, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0219, "step": 31, "tokens/total": 934400, "tokens/train_per_sec_per_gpu": 34.48, "tokens/trainable": 14064} +{"epoch": 0.125, "grad_norm": 0.8080421686172485, "learning_rate": 9.996629668474818e-05, "loss": 0.03089057095348835, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.03137, "step": 32, "tokens/total": 964832, "tokens/train_per_sec_per_gpu": 39.38, "tokens/trainable": 14565} +{"epoch": 0.12890625, "grad_norm": 1.3231370449066162, "learning_rate": 9.995412811123711e-05, "loss": 0.03202879801392555, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.03255, "step": 33, "tokens/total": 995040, "tokens/train_per_sec_per_gpu": 34.2, "tokens/trainable": 15005} +{"epoch": 0.1328125, "grad_norm": 1.8419537544250488, "learning_rate": 9.994008881295999e-05, "loss": 0.018571095541119576, "memory/device_reserved (GiB)": 34.72, "memory/max_active (GiB)": 33.64, "memory/max_allocated (GiB)": 33.64, "ppl": 1.01874, "step": 34, "tokens/total": 1024896, "tokens/train_per_sec_per_gpu": 32.47, "tokens/trainable": 15459} +{"epoch": 0.13671875, "grad_norm": 1.5461390018463135, "learning_rate": 9.992417937414932e-05, "loss": 0.03690098226070404, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.03759, "step": 35, "tokens/total": 1054992, "tokens/train_per_sec_per_gpu": 38.64, "tokens/trainable": 15960} +{"epoch": 0.140625, "grad_norm": 4.299174785614014, "learning_rate": 9.99064004568618e-05, "loss": 0.028481774032115936, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.02889, "step": 36, "tokens/total": 1085280, "tokens/train_per_sec_per_gpu": 34.98, "tokens/trainable": 16428} +{"epoch": 0.14453125, "grad_norm": 0.723717987537384, "learning_rate": 9.988675280095074e-05, "loss": 0.0157011728733778, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.01583, "step": 37, "tokens/total": 1115504, "tokens/train_per_sec_per_gpu": 32.18, "tokens/trainable": 16884} +{"epoch": 0.1484375, "grad_norm": 0.6181128025054932, "learning_rate": 9.986523722403528e-05, "loss": 0.02132808044552803, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02156, "step": 38, "tokens/total": 1145712, "tokens/train_per_sec_per_gpu": 32.34, "tokens/trainable": 17341} +{"epoch": 0.15234375, "grad_norm": 0.7512460350990295, "learning_rate": 9.984185462146642e-05, "loss": 0.023557491600513458, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.02384, "step": 39, "tokens/total": 1176128, "tokens/train_per_sec_per_gpu": 34.5, "tokens/trainable": 17786} +{"epoch": 0.15625, "grad_norm": 0.8244909644126892, "learning_rate": 9.98166059662897e-05, "loss": 0.031878214329481125, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.03239, "step": 40, "tokens/total": 1206576, "tokens/train_per_sec_per_gpu": 35.41, "tokens/trainable": 18262} +{"epoch": 0.16015625, "grad_norm": 0.5963007211685181, "learning_rate": 9.978949230920472e-05, "loss": 0.007491848897188902, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00752, "step": 41, "tokens/total": 1236848, "tokens/train_per_sec_per_gpu": 33.1, "tokens/trainable": 18716} +{"epoch": 0.1640625, "grad_norm": 1.3698993921279907, "learning_rate": 9.976051477852141e-05, "loss": 0.055056557059288025, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.0566, "step": 42, "tokens/total": 1267088, "tokens/train_per_sec_per_gpu": 33.28, "tokens/trainable": 19157} +{"epoch": 0.16796875, "grad_norm": 5.950209617614746, "learning_rate": 9.972967458011312e-05, "loss": 0.040029894560575485, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.04084, "step": 43, "tokens/total": 1297360, "tokens/train_per_sec_per_gpu": 36.88, "tokens/trainable": 19647} +{"epoch": 0.171875, "grad_norm": 0.45418253540992737, "learning_rate": 9.96969729973664e-05, "loss": 0.01683747209608555, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01698, "step": 44, "tokens/total": 1327744, "tokens/train_per_sec_per_gpu": 35.26, "tokens/trainable": 20119} +{"epoch": 0.17578125, "grad_norm": 0.2594800591468811, "learning_rate": 9.966241139112754e-05, "loss": 0.00699122529476881, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00702, "step": 45, "tokens/total": 1358096, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 20560} +{"epoch": 0.1796875, "grad_norm": 0.32979631423950195, "learning_rate": 9.96259911996461e-05, "loss": 0.007912673056125641, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00794, "step": 46, "tokens/total": 1388240, "tokens/train_per_sec_per_gpu": 32.98, "tokens/trainable": 20992} +{"epoch": 0.18359375, "grad_norm": 4.0735039710998535, "learning_rate": 9.958771393851491e-05, "loss": 0.03896857053041458, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.26, "memory/max_allocated (GiB)": 33.26, "ppl": 1.03974, "step": 47, "tokens/total": 1416240, "tokens/train_per_sec_per_gpu": 31.79, "tokens/trainable": 21441} +{"epoch": 0.1875, "grad_norm": 1.569137692451477, "learning_rate": 9.954758120060702e-05, "loss": 0.04643790051341057, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.04753, "step": 48, "tokens/total": 1446880, "tokens/train_per_sec_per_gpu": 33.82, "tokens/trainable": 21901} +{"epoch": 0.19140625, "grad_norm": 0.3084900975227356, "learning_rate": 9.950559465600948e-05, "loss": 0.005935921333730221, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00595, "step": 49, "tokens/total": 1477168, "tokens/train_per_sec_per_gpu": 33.62, "tokens/trainable": 22341} +{"epoch": 0.1953125, "grad_norm": 0.8096332550048828, "learning_rate": 9.946175605195379e-05, "loss": 0.015453304164111614, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01557, "step": 50, "tokens/total": 1507712, "tokens/train_per_sec_per_gpu": 33.5, "tokens/trainable": 22833} +{"epoch": 0.19921875, "grad_norm": 0.6452654004096985, "learning_rate": 9.941606721274322e-05, "loss": 0.012829513289034367, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01291, "step": 51, "tokens/total": 1537936, "tokens/train_per_sec_per_gpu": 31.01, "tokens/trainable": 23277} +{"epoch": 0.203125, "grad_norm": 1.566336750984192, "learning_rate": 9.936853003967685e-05, "loss": 0.027271470054984093, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02765, "step": 52, "tokens/total": 1568352, "tokens/train_per_sec_per_gpu": 36.02, "tokens/trainable": 23759} +{"epoch": 0.20703125, "grad_norm": 1.660810947418213, "learning_rate": 9.93191465109705e-05, "loss": 0.07711329311132431, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.08016, "step": 53, "tokens/total": 1598992, "tokens/train_per_sec_per_gpu": 33.3, "tokens/trainable": 24193} +{"epoch": 0.2109375, "grad_norm": 0.8398771286010742, "learning_rate": 9.926791868167438e-05, "loss": 0.02643466927111149, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.02679, "step": 54, "tokens/total": 1629488, "tokens/train_per_sec_per_gpu": 33.92, "tokens/trainable": 24619} +{"epoch": 0.21484375, "grad_norm": 1.306341290473938, "learning_rate": 9.921484868358753e-05, "loss": 0.01740848645567894, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01756, "step": 55, "tokens/total": 1659696, "tokens/train_per_sec_per_gpu": 33.19, "tokens/trainable": 25075} +{"epoch": 0.21875, "grad_norm": 0.6005977988243103, "learning_rate": 9.915993872516924e-05, "loss": 0.020611682906746864, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.02083, "step": 56, "tokens/total": 1689872, "tokens/train_per_sec_per_gpu": 31.86, "tokens/trainable": 25519} +{"epoch": 0.22265625, "grad_norm": 0.22082990407943726, "learning_rate": 9.9103191091447e-05, "loss": 0.00663959514349699, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00666, "step": 57, "tokens/total": 1720144, "tokens/train_per_sec_per_gpu": 33.02, "tokens/trainable": 25966} +{"epoch": 0.2265625, "grad_norm": 0.7464659810066223, "learning_rate": 9.904460814392147e-05, "loss": 0.015967492014169693, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0161, "step": 58, "tokens/total": 1750448, "tokens/train_per_sec_per_gpu": 34.25, "tokens/trainable": 26423} +{"epoch": 0.23046875, "grad_norm": 1.5382624864578247, "learning_rate": 9.898419232046825e-05, "loss": 0.02294602431356907, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.02321, "step": 59, "tokens/total": 1781088, "tokens/train_per_sec_per_gpu": 38.79, "tokens/trainable": 26934} +{"epoch": 0.234375, "grad_norm": 0.1427021324634552, "learning_rate": 9.892194613523633e-05, "loss": 0.004101241007447243, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00411, "step": 60, "tokens/total": 1811344, "tokens/train_per_sec_per_gpu": 33.36, "tokens/trainable": 27393} +{"epoch": 0.23828125, "grad_norm": 0.22952979803085327, "learning_rate": 9.885787217854357e-05, "loss": 0.008820527233183384, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00886, "step": 61, "tokens/total": 1841600, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 27852} +{"epoch": 0.2421875, "grad_norm": 0.5619235038757324, "learning_rate": 9.879197311676887e-05, "loss": 0.023235434666275978, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02351, "step": 62, "tokens/total": 1872048, "tokens/train_per_sec_per_gpu": 33.3, "tokens/trainable": 28292} +{"epoch": 0.24609375, "grad_norm": 0.5525861978530884, "learning_rate": 9.872425169224113e-05, "loss": 0.016968555748462677, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01711, "step": 63, "tokens/total": 1902592, "tokens/train_per_sec_per_gpu": 37.73, "tokens/trainable": 28798} +{"epoch": 0.25, "grad_norm": 2.4921607971191406, "learning_rate": 9.865471072312528e-05, "loss": 0.02083023451268673, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.02105, "step": 64, "tokens/total": 1933088, "tokens/train_per_sec_per_gpu": 36.11, "tokens/trainable": 29283} +{"epoch": 0.25390625, "grad_norm": 0.297742635011673, "learning_rate": 9.858335310330492e-05, "loss": 0.007790482137352228, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00782, "step": 65, "tokens/total": 1963296, "tokens/train_per_sec_per_gpu": 32.25, "tokens/trainable": 29745} +{"epoch": 0.2578125, "grad_norm": 0.6105188131332397, "learning_rate": 9.851018180226185e-05, "loss": 0.012852677144110203, "memory/device_reserved (GiB)": 34.87, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01294, "step": 66, "tokens/total": 1993760, "tokens/train_per_sec_per_gpu": 31.48, "tokens/trainable": 30186} +{"epoch": 0.26171875, "grad_norm": 0.542284369468689, "learning_rate": 9.843519986495259e-05, "loss": 0.004466134589165449, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00448, "step": 67, "tokens/total": 2024144, "tokens/train_per_sec_per_gpu": 33.37, "tokens/trainable": 30622} +{"epoch": 0.265625, "grad_norm": 0.32131820917129517, "learning_rate": 9.835841041168162e-05, "loss": 0.009684666059911251, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00973, "step": 68, "tokens/total": 2054608, "tokens/train_per_sec_per_gpu": 35.74, "tokens/trainable": 31097} +{"epoch": 0.26953125, "grad_norm": 0.5698699951171875, "learning_rate": 9.82798166379715e-05, "loss": 0.01115076057612896, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01121, "step": 69, "tokens/total": 2084912, "tokens/train_per_sec_per_gpu": 36.87, "tokens/trainable": 31595} +{"epoch": 0.2734375, "grad_norm": 0.3763757348060608, "learning_rate": 9.819942181443002e-05, "loss": 0.004509727004915476, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00452, "step": 70, "tokens/total": 2115216, "tokens/train_per_sec_per_gpu": 31.0, "tokens/trainable": 32036} +{"epoch": 0.27734375, "grad_norm": 0.9681234359741211, "learning_rate": 9.811722928661392e-05, "loss": 0.02017754502594471, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02038, "step": 71, "tokens/total": 2145424, "tokens/train_per_sec_per_gpu": 28.34, "tokens/trainable": 32428} +{"epoch": 0.28125, "grad_norm": 0.2625149190425873, "learning_rate": 9.803324247488975e-05, "loss": 0.002312284428626299, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00231, "step": 72, "tokens/total": 2175648, "tokens/train_per_sec_per_gpu": 32.44, "tokens/trainable": 32880} +{"epoch": 0.28515625, "grad_norm": 0.3928808569908142, "learning_rate": 9.794746487429161e-05, "loss": 0.0031382141169160604, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00314, "step": 73, "tokens/total": 2205856, "tokens/train_per_sec_per_gpu": 33.88, "tokens/trainable": 33323} +{"epoch": 0.2890625, "grad_norm": 0.9950233697891235, "learning_rate": 9.785990005437554e-05, "loss": 0.026497021317481995, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.02685, "step": 74, "tokens/total": 2236352, "tokens/train_per_sec_per_gpu": 36.17, "tokens/trainable": 33792} +{"epoch": 0.29296875, "grad_norm": 0.8722730278968811, "learning_rate": 9.777055165907117e-05, "loss": 0.01632213592529297, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01646, "step": 75, "tokens/total": 2266480, "tokens/train_per_sec_per_gpu": 35.6, "tokens/trainable": 34223} +{"epoch": 0.296875, "grad_norm": 1.1537584066390991, "learning_rate": 9.767942340652993e-05, "loss": 0.012922817841172218, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.01301, "step": 76, "tokens/total": 2296496, "tokens/train_per_sec_per_gpu": 29.9, "tokens/trainable": 34649} +{"epoch": 0.30078125, "grad_norm": 0.6076233386993408, "learning_rate": 9.758651908897035e-05, "loss": 0.016752732917666435, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01689, "step": 77, "tokens/total": 2327040, "tokens/train_per_sec_per_gpu": 35.7, "tokens/trainable": 35094} +{"epoch": 0.3046875, "grad_norm": 1.038780689239502, "learning_rate": 9.749184257252033e-05, "loss": 0.034519392997026443, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.03512, "step": 78, "tokens/total": 2357328, "tokens/train_per_sec_per_gpu": 31.97, "tokens/trainable": 35534} +{"epoch": 0.30859375, "grad_norm": 0.4420571029186249, "learning_rate": 9.739539779705614e-05, "loss": 0.021940411999821663, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.02218, "step": 79, "tokens/total": 2387664, "tokens/train_per_sec_per_gpu": 36.09, "tokens/trainable": 36029} +{"epoch": 0.3125, "grad_norm": 0.33440592885017395, "learning_rate": 9.729718877603861e-05, "loss": 0.007664849050343037, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00769, "step": 80, "tokens/total": 2418000, "tokens/train_per_sec_per_gpu": 35.84, "tokens/trainable": 36469} +{"epoch": 0.31640625, "grad_norm": 0.8483979105949402, "learning_rate": 9.719721959634592e-05, "loss": 0.04067182168364525, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.04151, "step": 81, "tokens/total": 2448720, "tokens/train_per_sec_per_gpu": 31.05, "tokens/trainable": 36906} +{"epoch": 0.3203125, "grad_norm": 2.5027475357055664, "learning_rate": 9.709549441810375e-05, "loss": 0.012142607942223549, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.01222, "step": 82, "tokens/total": 2479248, "tokens/train_per_sec_per_gpu": 39.28, "tokens/trainable": 37390} +{"epoch": 0.32421875, "grad_norm": 0.25906872749328613, "learning_rate": 9.699201747451195e-05, "loss": 0.008427562192082405, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00846, "step": 83, "tokens/total": 2509408, "tokens/train_per_sec_per_gpu": 31.11, "tokens/trainable": 37838} +{"epoch": 0.328125, "grad_norm": 0.5375878214836121, "learning_rate": 9.688679307166854e-05, "loss": 0.010640576481819153, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0107, "step": 84, "tokens/total": 2539776, "tokens/train_per_sec_per_gpu": 35.48, "tokens/trainable": 38310} +{"epoch": 0.33203125, "grad_norm": 1.079277753829956, "learning_rate": 9.677982558839042e-05, "loss": 0.0304670799523592, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.03094, "step": 85, "tokens/total": 2569840, "tokens/train_per_sec_per_gpu": 34.4, "tokens/trainable": 38789} +{"epoch": 0.3359375, "grad_norm": 0.3869364857673645, "learning_rate": 9.66711194760312e-05, "loss": 0.015695516020059586, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01582, "step": 86, "tokens/total": 2600192, "tokens/train_per_sec_per_gpu": 36.71, "tokens/trainable": 39277} +{"epoch": 0.33984375, "grad_norm": 0.427566796541214, "learning_rate": 9.656067925829593e-05, "loss": 0.00836772657930851, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0084, "step": 87, "tokens/total": 2630640, "tokens/train_per_sec_per_gpu": 35.9, "tokens/trainable": 39737} +{"epoch": 0.34375, "grad_norm": 0.4263860881328583, "learning_rate": 9.644850953105288e-05, "loss": 0.00825466588139534, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00829, "step": 88, "tokens/total": 2660832, "tokens/train_per_sec_per_gpu": 29.8, "tokens/trainable": 40179} +{"epoch": 0.34765625, "grad_norm": 1.1659919023513794, "learning_rate": 9.633461496214225e-05, "loss": 0.01808004267513752, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01824, "step": 89, "tokens/total": 2691328, "tokens/train_per_sec_per_gpu": 32.88, "tokens/trainable": 40649} +{"epoch": 0.3515625, "grad_norm": 0.5744115114212036, "learning_rate": 9.621900029118195e-05, "loss": 0.021331775933504105, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.41, "memory/max_allocated (GiB)": 33.41, "ppl": 1.02156, "step": 90, "tokens/total": 2719696, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 41080} +{"epoch": 0.35546875, "grad_norm": 0.26662108302116394, "learning_rate": 9.610167032937036e-05, "loss": 0.0042778924107551575, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00429, "step": 91, "tokens/total": 2750272, "tokens/train_per_sec_per_gpu": 38.35, "tokens/trainable": 41599} +{"epoch": 0.359375, "grad_norm": 0.48243093490600586, "learning_rate": 9.598262995928611e-05, "loss": 0.01029058825224638, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.01034, "step": 92, "tokens/total": 2780656, "tokens/train_per_sec_per_gpu": 37.08, "tokens/trainable": 42076} +{"epoch": 0.36328125, "grad_norm": 0.3396889865398407, "learning_rate": 9.586188413468492e-05, "loss": 0.008061517961323261, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00809, "step": 93, "tokens/total": 2811088, "tokens/train_per_sec_per_gpu": 34.0, "tokens/trainable": 42522} +{"epoch": 0.3671875, "grad_norm": 0.7351086139678955, "learning_rate": 9.57394378802934e-05, "loss": 0.02914494089782238, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02957, "step": 94, "tokens/total": 2841520, "tokens/train_per_sec_per_gpu": 34.69, "tokens/trainable": 42974} +{"epoch": 0.37109375, "grad_norm": 1.0318830013275146, "learning_rate": 9.56152962916e-05, "loss": 0.027875451371073723, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.02827, "step": 95, "tokens/total": 2871600, "tokens/train_per_sec_per_gpu": 31.0, "tokens/trainable": 43380} +{"epoch": 0.375, "grad_norm": 0.4948607385158539, "learning_rate": 9.548946453464296e-05, "loss": 0.01107320748269558, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01113, "step": 96, "tokens/total": 2902144, "tokens/train_per_sec_per_gpu": 34.38, "tokens/trainable": 43865} +{"epoch": 0.37890625, "grad_norm": 0.1577538549900055, "learning_rate": 9.53619478457953e-05, "loss": 0.006561641581356525, "memory/device_reserved (GiB)": 36.1, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00658, "step": 97, "tokens/total": 2932272, "tokens/train_per_sec_per_gpu": 32.26, "tokens/trainable": 44328} +{"epoch": 0.3828125, "grad_norm": 0.2749426066875458, "learning_rate": 9.523275153154695e-05, "loss": 0.007920589298009872, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00795, "step": 98, "tokens/total": 2962032, "tokens/train_per_sec_per_gpu": 31.3, "tokens/trainable": 44778} +{"epoch": 0.38671875, "grad_norm": 0.47395026683807373, "learning_rate": 9.51018809682839e-05, "loss": 0.03451812267303467, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.03512, "step": 99, "tokens/total": 2992256, "tokens/train_per_sec_per_gpu": 37.58, "tokens/trainable": 45225} +{"epoch": 0.390625, "grad_norm": 0.5378239750862122, "learning_rate": 9.49693416020645e-05, "loss": 0.017406146973371506, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01756, "step": 100, "tokens/total": 3022608, "tokens/train_per_sec_per_gpu": 36.15, "tokens/trainable": 45678} +{"epoch": 0.39453125, "grad_norm": 0.45281967520713806, "learning_rate": 9.483513894839276e-05, "loss": 0.004519028589129448, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00453, "step": 101, "tokens/total": 3052992, "tokens/train_per_sec_per_gpu": 32.99, "tokens/trainable": 46125} +{"epoch": 0.3984375, "grad_norm": 1.7491872310638428, "learning_rate": 9.469927859198888e-05, "loss": 0.00577917555347085, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.0058, "step": 102, "tokens/total": 3083392, "tokens/train_per_sec_per_gpu": 40.05, "tokens/trainable": 46612} +{"epoch": 0.40234375, "grad_norm": 0.3060601055622101, "learning_rate": 9.456176618655689e-05, "loss": 0.008232416585087776, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00827, "step": 103, "tokens/total": 3113744, "tokens/train_per_sec_per_gpu": 32.96, "tokens/trainable": 47059} +{"epoch": 0.40625, "grad_norm": 0.7810171246528625, "learning_rate": 9.442260745454927e-05, "loss": 0.03279104456305504, "memory/device_reserved (GiB)": 35.8, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.03333, "step": 104, "tokens/total": 3144272, "tokens/train_per_sec_per_gpu": 34.4, "tokens/trainable": 47536} +{"epoch": 0.41015625, "grad_norm": 0.4873018264770508, "learning_rate": 9.428180818692884e-05, "loss": 0.008343804627656937, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00838, "step": 105, "tokens/total": 3174512, "tokens/train_per_sec_per_gpu": 34.4, "tokens/trainable": 48037} +{"epoch": 0.4140625, "grad_norm": 0.26075872778892517, "learning_rate": 9.413937424292791e-05, "loss": 0.00894663855433464, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00899, "step": 106, "tokens/total": 3204896, "tokens/train_per_sec_per_gpu": 32.94, "tokens/trainable": 48491} +{"epoch": 0.41796875, "grad_norm": 0.9285218715667725, "learning_rate": 9.399531154980424e-05, "loss": 0.0386667475104332, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.03942, "step": 107, "tokens/total": 3235360, "tokens/train_per_sec_per_gpu": 33.31, "tokens/trainable": 48940} +{"epoch": 0.421875, "grad_norm": 0.37687361240386963, "learning_rate": 9.384962610259455e-05, "loss": 0.014155298471450806, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01426, "step": 108, "tokens/total": 3265552, "tokens/train_per_sec_per_gpu": 32.06, "tokens/trainable": 49389} +{"epoch": 0.42578125, "grad_norm": 0.12366248667240143, "learning_rate": 9.370232396386494e-05, "loss": 0.0033611906692385674, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.00337, "step": 109, "tokens/total": 3293856, "tokens/train_per_sec_per_gpu": 36.26, "tokens/trainable": 49838} +{"epoch": 0.4296875, "grad_norm": 0.3310869038105011, "learning_rate": 9.355341126345868e-05, "loss": 0.01677614450454712, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.01692, "step": 110, "tokens/total": 3323984, "tokens/train_per_sec_per_gpu": 34.17, "tokens/trainable": 50310} +{"epoch": 0.43359375, "grad_norm": 0.16138073801994324, "learning_rate": 9.340289419824107e-05, "loss": 0.005432053469121456, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00545, "step": 111, "tokens/total": 3354320, "tokens/train_per_sec_per_gpu": 33.94, "tokens/trainable": 50755} +{"epoch": 0.4375, "grad_norm": 0.2602401673793793, "learning_rate": 9.325077903184159e-05, "loss": 0.009463014081120491, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00951, "step": 112, "tokens/total": 3384880, "tokens/train_per_sec_per_gpu": 31.85, "tokens/trainable": 51213} +{"epoch": 0.44140625, "grad_norm": 0.11247953772544861, "learning_rate": 9.30970720943932e-05, "loss": 0.0028001072350889444, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0028, "step": 113, "tokens/total": 3415104, "tokens/train_per_sec_per_gpu": 34.14, "tokens/trainable": 51660} +{"epoch": 0.4453125, "grad_norm": 0.2779226303100586, "learning_rate": 9.2941779782269e-05, "loss": 0.01312212087213993, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01321, "step": 114, "tokens/total": 3445504, "tokens/train_per_sec_per_gpu": 32.72, "tokens/trainable": 52133} +{"epoch": 0.44921875, "grad_norm": 0.22592218220233917, "learning_rate": 9.278490855781596e-05, "loss": 0.005544666200876236, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00556, "step": 115, "tokens/total": 3475744, "tokens/train_per_sec_per_gpu": 37.85, "tokens/trainable": 52580} +{"epoch": 0.453125, "grad_norm": 0.272401362657547, "learning_rate": 9.262646494908604e-05, "loss": 0.005348580423742533, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00536, "step": 116, "tokens/total": 3506016, "tokens/train_per_sec_per_gpu": 35.2, "tokens/trainable": 53033} +{"epoch": 0.45703125, "grad_norm": 0.9178272485733032, "learning_rate": 9.246645554956457e-05, "loss": 0.022306114435195923, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.02256, "step": 117, "tokens/total": 3536256, "tokens/train_per_sec_per_gpu": 35.66, "tokens/trainable": 53517} +{"epoch": 0.4609375, "grad_norm": 0.10259439796209335, "learning_rate": 9.230488701789578e-05, "loss": 0.0009993483545258641, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.001, "step": 118, "tokens/total": 3566480, "tokens/train_per_sec_per_gpu": 34.96, "tokens/trainable": 53967} +{"epoch": 0.46484375, "grad_norm": 0.15369175374507904, "learning_rate": 9.214176607760577e-05, "loss": 0.0015791363548487425, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00158, "step": 119, "tokens/total": 3596624, "tokens/train_per_sec_per_gpu": 35.68, "tokens/trainable": 54430} +{"epoch": 0.46875, "grad_norm": 0.4007995128631592, "learning_rate": 9.197709951682268e-05, "loss": 0.008637797087430954, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00868, "step": 120, "tokens/total": 3627168, "tokens/train_per_sec_per_gpu": 35.57, "tokens/trainable": 54896} +{"epoch": 0.47265625, "grad_norm": 2.9844160079956055, "learning_rate": 9.181089418799428e-05, "loss": 0.01763911172747612, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0178, "step": 121, "tokens/total": 3657632, "tokens/train_per_sec_per_gpu": 38.12, "tokens/trainable": 55379} +{"epoch": 0.4765625, "grad_norm": 0.4593406021595001, "learning_rate": 9.164315700760271e-05, "loss": 0.007832051254808903, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00786, "step": 122, "tokens/total": 3687824, "tokens/train_per_sec_per_gpu": 32.05, "tokens/trainable": 55803} +{"epoch": 0.48046875, "grad_norm": 0.22288177907466888, "learning_rate": 9.147389495587671e-05, "loss": 0.0024647461250424385, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00247, "step": 123, "tokens/total": 3718320, "tokens/train_per_sec_per_gpu": 36.8, "tokens/trainable": 56249} +{"epoch": 0.484375, "grad_norm": 3.4173779487609863, "learning_rate": 9.130311507650116e-05, "loss": 0.010984445922076702, "memory/device_reserved (GiB)": 37.62, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01104, "step": 124, "tokens/total": 3748672, "tokens/train_per_sec_per_gpu": 32.77, "tokens/trainable": 56713} +{"epoch": 0.48828125, "grad_norm": 0.46252161264419556, "learning_rate": 9.113082447632394e-05, "loss": 0.0051902830600738525, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0052, "step": 125, "tokens/total": 3778976, "tokens/train_per_sec_per_gpu": 39.56, "tokens/trainable": 57196} +{"epoch": 0.4921875, "grad_norm": 0.3998625874519348, "learning_rate": 9.09570303250602e-05, "loss": 0.004260494839400053, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00427, "step": 126, "tokens/total": 3809296, "tokens/train_per_sec_per_gpu": 36.93, "tokens/trainable": 57680} +{"epoch": 0.49609375, "grad_norm": 0.23200765252113342, "learning_rate": 9.078173985499394e-05, "loss": 0.003244318999350071, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00325, "step": 127, "tokens/total": 3839328, "tokens/train_per_sec_per_gpu": 32.22, "tokens/trainable": 58110} +{"epoch": 0.5, "grad_norm": 0.26066046953201294, "learning_rate": 9.060496036067713e-05, "loss": 0.011201048269867897, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01126, "step": 128, "tokens/total": 3869824, "tokens/train_per_sec_per_gpu": 32.02, "tokens/trainable": 58534} +{"epoch": 0.50390625, "grad_norm": 0.7281484007835388, "learning_rate": 9.042669919862615e-05, "loss": 0.011813497170805931, "memory/device_reserved (GiB)": 37.76, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01188, "step": 129, "tokens/total": 3900080, "tokens/train_per_sec_per_gpu": 35.31, "tokens/trainable": 58998} +{"epoch": 0.5078125, "grad_norm": 0.5487067699432373, "learning_rate": 9.024696378701557e-05, "loss": 0.008090369403362274, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00812, "step": 130, "tokens/total": 3930560, "tokens/train_per_sec_per_gpu": 32.06, "tokens/trainable": 59448} +{"epoch": 0.51171875, "grad_norm": 0.5761887431144714, "learning_rate": 9.006576160536948e-05, "loss": 0.008387972600758076, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00842, "step": 131, "tokens/total": 3960496, "tokens/train_per_sec_per_gpu": 32.4, "tokens/trainable": 59906} +{"epoch": 0.515625, "grad_norm": 0.34462451934814453, "learning_rate": 8.988310019425035e-05, "loss": 0.005629569757729769, "memory/device_reserved (GiB)": 35.93, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00565, "step": 132, "tokens/total": 3990928, "tokens/train_per_sec_per_gpu": 34.59, "tokens/trainable": 60350} +{"epoch": 0.51953125, "grad_norm": 0.7096816897392273, "learning_rate": 8.969898715494506e-05, "loss": 0.010645382106304169, "memory/device_reserved (GiB)": 37.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0107, "step": 133, "tokens/total": 4021264, "tokens/train_per_sec_per_gpu": 36.48, "tokens/trainable": 60831} +{"epoch": 0.5234375, "grad_norm": 0.9463919997215271, "learning_rate": 8.951343014914869e-05, "loss": 0.018044205382466316, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01821, "step": 134, "tokens/total": 4051728, "tokens/train_per_sec_per_gpu": 34.34, "tokens/trainable": 61305} +{"epoch": 0.52734375, "grad_norm": 0.20147432386875153, "learning_rate": 8.932643689864568e-05, "loss": 0.00538910087198019, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.0054, "step": 135, "tokens/total": 4081920, "tokens/train_per_sec_per_gpu": 38.01, "tokens/trainable": 61792} +{"epoch": 0.53125, "grad_norm": 0.1833702027797699, "learning_rate": 8.913801518498845e-05, "loss": 0.00374586577527225, "memory/device_reserved (GiB)": 38.55, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00375, "step": 136, "tokens/total": 4112304, "tokens/train_per_sec_per_gpu": 31.67, "tokens/trainable": 62253} +{"epoch": 0.53515625, "grad_norm": 0.7401099801063538, "learning_rate": 8.894817284917364e-05, "loss": 0.01421094499528408, "memory/device_reserved (GiB)": 38.55, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01431, "step": 137, "tokens/total": 4142512, "tokens/train_per_sec_per_gpu": 31.97, "tokens/trainable": 62707} +{"epoch": 0.5390625, "grad_norm": 0.2769958972930908, "learning_rate": 8.875691779131569e-05, "loss": 0.0012122916523367167, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00121, "step": 138, "tokens/total": 4172992, "tokens/train_per_sec_per_gpu": 29.6, "tokens/trainable": 63145} +{"epoch": 0.54296875, "grad_norm": 0.2656077444553375, "learning_rate": 8.856425797031829e-05, "loss": 0.005337266717106104, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00535, "step": 139, "tokens/total": 4203136, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 63587} +{"epoch": 0.546875, "grad_norm": 0.15931734442710876, "learning_rate": 8.837020140354295e-05, "loss": 0.003361564828082919, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00337, "step": 140, "tokens/total": 4233456, "tokens/train_per_sec_per_gpu": 36.77, "tokens/trainable": 64052} +{"epoch": 0.55078125, "grad_norm": 0.2797258496284485, "learning_rate": 8.817475616647554e-05, "loss": 0.006030731834471226, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00605, "step": 141, "tokens/total": 4263728, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 64526} +{"epoch": 0.5546875, "grad_norm": 0.18650899827480316, "learning_rate": 8.797793039239017e-05, "loss": 0.0034353849478065968, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.00344, "step": 142, "tokens/total": 4294432, "tokens/train_per_sec_per_gpu": 34.97, "tokens/trainable": 64983} +{"epoch": 0.55859375, "grad_norm": 0.18854950368404388, "learning_rate": 8.777973227201069e-05, "loss": 0.004699597600847483, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00471, "step": 143, "tokens/total": 4324960, "tokens/train_per_sec_per_gpu": 38.02, "tokens/trainable": 65492} +{"epoch": 0.5625, "grad_norm": 0.1250438392162323, "learning_rate": 8.758017005316988e-05, "loss": 0.0013500247150659561, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00135, "step": 144, "tokens/total": 4355424, "tokens/train_per_sec_per_gpu": 36.06, "tokens/trainable": 65925} +{"epoch": 0.56640625, "grad_norm": 0.2933589518070221, "learning_rate": 8.737925204046629e-05, "loss": 0.008707849308848381, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00875, "step": 145, "tokens/total": 4385712, "tokens/train_per_sec_per_gpu": 31.73, "tokens/trainable": 66383} +{"epoch": 0.5703125, "grad_norm": 0.30792802572250366, "learning_rate": 8.717698659491851e-05, "loss": 0.005249719135463238, "memory/device_reserved (GiB)": 38.56, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00526, "step": 146, "tokens/total": 4416016, "tokens/train_per_sec_per_gpu": 31.65, "tokens/trainable": 66840} +{"epoch": 0.57421875, "grad_norm": 1.029360055923462, "learning_rate": 8.697338213361735e-05, "loss": 0.014694154262542725, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0148, "step": 147, "tokens/total": 4446368, "tokens/train_per_sec_per_gpu": 32.05, "tokens/trainable": 67297} +{"epoch": 0.578125, "grad_norm": 0.2955895662307739, "learning_rate": 8.676844712937552e-05, "loss": 0.004935478325933218, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00495, "step": 148, "tokens/total": 4476880, "tokens/train_per_sec_per_gpu": 37.63, "tokens/trainable": 67769} +{"epoch": 0.58203125, "grad_norm": 0.556270956993103, "learning_rate": 8.656219011037509e-05, "loss": 0.023344792425632477, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.02362, "step": 149, "tokens/total": 4507232, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 68257} +{"epoch": 0.5859375, "grad_norm": 0.43625330924987793, "learning_rate": 8.63546196598125e-05, "loss": 0.009602912701666355, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00965, "step": 150, "tokens/total": 4537376, "tokens/train_per_sec_per_gpu": 35.86, "tokens/trainable": 68706} +{"epoch": 0.58984375, "grad_norm": 0.2960459291934967, "learning_rate": 8.614574441554145e-05, "loss": 0.009531187824904919, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00958, "step": 151, "tokens/total": 4567584, "tokens/train_per_sec_per_gpu": 33.61, "tokens/trainable": 69131} +{"epoch": 0.59375, "grad_norm": 0.22226394712924957, "learning_rate": 8.593557306971349e-05, "loss": 0.013231074437499046, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.01332, "step": 152, "tokens/total": 4597792, "tokens/train_per_sec_per_gpu": 32.24, "tokens/trainable": 69586} +{"epoch": 0.59765625, "grad_norm": 0.13233518600463867, "learning_rate": 8.572411436841618e-05, "loss": 0.0031906149815768003, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.0032, "step": 153, "tokens/total": 4627936, "tokens/train_per_sec_per_gpu": 33.11, "tokens/trainable": 70061} +{"epoch": 0.6015625, "grad_norm": 0.17915832996368408, "learning_rate": 8.551137711130922e-05, "loss": 0.006568522658199072, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00659, "step": 154, "tokens/total": 4658352, "tokens/train_per_sec_per_gpu": 28.23, "tokens/trainable": 70467} +{"epoch": 0.60546875, "grad_norm": 0.3126574456691742, "learning_rate": 8.529737015125824e-05, "loss": 0.014315147884190083, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.01442, "step": 155, "tokens/total": 4688896, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 70933} +{"epoch": 0.609375, "grad_norm": 0.19957849383354187, "learning_rate": 8.508210239396639e-05, "loss": 0.0077377320267260075, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00777, "step": 156, "tokens/total": 4719168, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 71382} +{"epoch": 0.61328125, "grad_norm": 0.07674852013587952, "learning_rate": 8.486558279760375e-05, "loss": 0.0022812150418758392, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00228, "step": 157, "tokens/total": 4749136, "tokens/train_per_sec_per_gpu": 31.02, "tokens/trainable": 71808} +{"epoch": 0.6171875, "grad_norm": 0.29424574971199036, "learning_rate": 8.464782037243449e-05, "loss": 0.009841009974479675, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00989, "step": 158, "tokens/total": 4779472, "tokens/train_per_sec_per_gpu": 33.19, "tokens/trainable": 72249} +{"epoch": 0.62109375, "grad_norm": 0.38520634174346924, "learning_rate": 8.442882418044202e-05, "loss": 0.017565667629241943, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.01772, "step": 159, "tokens/total": 4809632, "tokens/train_per_sec_per_gpu": 35.54, "tokens/trainable": 72726} +{"epoch": 0.625, "grad_norm": 0.1961541771888733, "learning_rate": 8.420860333495179e-05, "loss": 0.00602555088698864, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00604, "step": 160, "tokens/total": 4840112, "tokens/train_per_sec_per_gpu": 33.07, "tokens/trainable": 73148} +{"epoch": 0.62890625, "grad_norm": 0.3786018490791321, "learning_rate": 8.398716700025208e-05, "loss": 0.0033199097961187363, "memory/device_reserved (GiB)": 38.65, "memory/max_active (GiB)": 34.04, "memory/max_allocated (GiB)": 34.04, "ppl": 1.00333, "step": 161, "tokens/total": 4870656, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 73600} +{"epoch": 0.6328125, "grad_norm": 1.5304594039916992, "learning_rate": 8.376452439121266e-05, "loss": 0.013961974531412125, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.01406, "step": 162, "tokens/total": 4900608, "tokens/train_per_sec_per_gpu": 36.06, "tokens/trainable": 74068} +{"epoch": 0.63671875, "grad_norm": 0.3360394835472107, "learning_rate": 8.354068477290124e-05, "loss": 0.010342447087168694, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.0104, "step": 163, "tokens/total": 4930752, "tokens/train_per_sec_per_gpu": 36.1, "tokens/trainable": 74527} +{"epoch": 0.640625, "grad_norm": 0.1671362817287445, "learning_rate": 8.331565746019807e-05, "loss": 0.006741576828062534, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00676, "step": 164, "tokens/total": 4961008, "tokens/train_per_sec_per_gpu": 32.16, "tokens/trainable": 74992} +{"epoch": 0.64453125, "grad_norm": 0.338470995426178, "learning_rate": 8.308945181740812e-05, "loss": 0.009520573541522026, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00957, "step": 165, "tokens/total": 4991200, "tokens/train_per_sec_per_gpu": 35.65, "tokens/trainable": 75442} +{"epoch": 0.6484375, "grad_norm": 0.06734886765480042, "learning_rate": 8.286207725787153e-05, "loss": 0.0011861161328852177, "memory/device_reserved (GiB)": 36.0, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00119, "step": 166, "tokens/total": 5021600, "tokens/train_per_sec_per_gpu": 31.54, "tokens/trainable": 75887} +{"epoch": 0.65234375, "grad_norm": 0.5632230043411255, "learning_rate": 8.263354324357182e-05, "loss": 0.02382841520011425, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.02411, "step": 167, "tokens/total": 5052032, "tokens/train_per_sec_per_gpu": 34.45, "tokens/trainable": 76331} +{"epoch": 0.65625, "grad_norm": 0.14913584291934967, "learning_rate": 8.240385928474219e-05, "loss": 0.002766846213489771, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00277, "step": 168, "tokens/total": 5080256, "tokens/train_per_sec_per_gpu": 36.19, "tokens/trainable": 76745} +{"epoch": 0.66015625, "grad_norm": 0.1904655545949936, "learning_rate": 8.217303493946967e-05, "loss": 0.0046575069427490234, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00467, "step": 169, "tokens/total": 5110784, "tokens/train_per_sec_per_gpu": 34.56, "tokens/trainable": 77201} +{"epoch": 0.6640625, "grad_norm": 0.3416374921798706, "learning_rate": 8.194107981329746e-05, "loss": 0.007014934439212084, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00704, "step": 170, "tokens/total": 5141200, "tokens/train_per_sec_per_gpu": 33.74, "tokens/trainable": 77655} +{"epoch": 0.66796875, "grad_norm": 0.06584338843822479, "learning_rate": 8.170800355882518e-05, "loss": 0.0010388410883024335, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00104, "step": 171, "tokens/total": 5171760, "tokens/train_per_sec_per_gpu": 39.64, "tokens/trainable": 78122} +{"epoch": 0.671875, "grad_norm": 1.4229398965835571, "learning_rate": 8.147381587530713e-05, "loss": 0.015414456836879253, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01553, "step": 172, "tokens/total": 5202272, "tokens/train_per_sec_per_gpu": 33.89, "tokens/trainable": 78576} +{"epoch": 0.67578125, "grad_norm": 0.3806024193763733, "learning_rate": 8.123852650824877e-05, "loss": 0.012036355212330818, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.01211, "step": 173, "tokens/total": 5232800, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 79059} +{"epoch": 0.6796875, "grad_norm": 0.5877760052680969, "learning_rate": 8.100214524900103e-05, "loss": 0.01602815091609955, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.01616, "step": 174, "tokens/total": 5262672, "tokens/train_per_sec_per_gpu": 29.99, "tokens/trainable": 79498} +{"epoch": 0.68359375, "grad_norm": 0.33037039637565613, "learning_rate": 8.076468193435301e-05, "loss": 0.010002637282013893, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01005, "step": 175, "tokens/total": 5293072, "tokens/train_per_sec_per_gpu": 30.73, "tokens/trainable": 79922} +{"epoch": 0.6875, "grad_norm": 0.4363207221031189, "learning_rate": 8.052614644612253e-05, "loss": 0.016971714794635773, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.46, "memory/max_allocated (GiB)": 33.46, "ppl": 1.01712, "step": 176, "tokens/total": 5321520, "tokens/train_per_sec_per_gpu": 36.01, "tokens/trainable": 80383} +{"epoch": 0.69140625, "grad_norm": 0.40556958317756653, "learning_rate": 8.028654871074489e-05, "loss": 0.00855817086994648, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00859, "step": 177, "tokens/total": 5351904, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 80824} +{"epoch": 0.6953125, "grad_norm": 0.16822558641433716, "learning_rate": 8.004589869885986e-05, "loss": 0.006781051866710186, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0068, "step": 178, "tokens/total": 5382432, "tokens/train_per_sec_per_gpu": 32.6, "tokens/trainable": 81243} +{"epoch": 0.69921875, "grad_norm": 0.28983059525489807, "learning_rate": 7.980420642489674e-05, "loss": 0.005999822169542313, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00602, "step": 179, "tokens/total": 5412960, "tokens/train_per_sec_per_gpu": 34.67, "tokens/trainable": 81716} +{"epoch": 0.703125, "grad_norm": 0.18040503561496735, "learning_rate": 7.95614819466576e-05, "loss": 0.004937571007758379, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00495, "step": 180, "tokens/total": 5443232, "tokens/train_per_sec_per_gpu": 35.67, "tokens/trainable": 82181} +{"epoch": 0.70703125, "grad_norm": 0.3842947483062744, "learning_rate": 7.931773536489872e-05, "loss": 0.014881981536746025, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.29, "memory/max_allocated (GiB)": 33.29, "ppl": 1.01499, "step": 181, "tokens/total": 5471440, "tokens/train_per_sec_per_gpu": 34.94, "tokens/trainable": 82626} +{"epoch": 0.7109375, "grad_norm": 0.1560991108417511, "learning_rate": 7.907297682291035e-05, "loss": 0.004242315422743559, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00425, "step": 182, "tokens/total": 5501744, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 83089} +{"epoch": 0.71484375, "grad_norm": 0.1657864898443222, "learning_rate": 7.882721650609442e-05, "loss": 0.005256622098386288, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00527, "step": 183, "tokens/total": 5532272, "tokens/train_per_sec_per_gpu": 35.64, "tokens/trainable": 83590} +{"epoch": 0.71875, "grad_norm": 0.6097539067268372, "learning_rate": 7.85804646415409e-05, "loss": 0.016960084438323975, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0171, "step": 184, "tokens/total": 5562784, "tokens/train_per_sec_per_gpu": 35.62, "tokens/trainable": 84046} +{"epoch": 0.72265625, "grad_norm": 0.09047499299049377, "learning_rate": 7.833273149760207e-05, "loss": 0.002440867479890585, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00244, "step": 185, "tokens/total": 5592800, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 84517} +{"epoch": 0.7265625, "grad_norm": 3.610114574432373, "learning_rate": 7.808402738346527e-05, "loss": 0.027849921956658363, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02824, "step": 186, "tokens/total": 5623088, "tokens/train_per_sec_per_gpu": 34.34, "tokens/trainable": 84974} +{"epoch": 0.73046875, "grad_norm": 0.09014768898487091, "learning_rate": 7.783436264872382e-05, "loss": 0.0016205484280362725, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00162, "step": 187, "tokens/total": 5653344, "tokens/train_per_sec_per_gpu": 35.48, "tokens/trainable": 85460} +{"epoch": 0.734375, "grad_norm": 0.11981519311666489, "learning_rate": 7.758374768294647e-05, "loss": 0.004210016690194607, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00422, "step": 188, "tokens/total": 5683600, "tokens/train_per_sec_per_gpu": 36.47, "tokens/trainable": 85939} +{"epoch": 0.73828125, "grad_norm": 0.21320489048957825, "learning_rate": 7.733219291524489e-05, "loss": 0.0067794108763337135, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0068, "step": 189, "tokens/total": 5711952, "tokens/train_per_sec_per_gpu": 38.83, "tokens/trainable": 86377} +{"epoch": 0.7421875, "grad_norm": 0.07177300751209259, "learning_rate": 7.707970881383977e-05, "loss": 0.0022305003367364407, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00223, "step": 190, "tokens/total": 5742336, "tokens/train_per_sec_per_gpu": 34.26, "tokens/trainable": 86834} +{"epoch": 0.74609375, "grad_norm": 0.8523200154304504, "learning_rate": 7.682630588562518e-05, "loss": 0.00609443848952651, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00611, "step": 191, "tokens/total": 5772560, "tokens/train_per_sec_per_gpu": 32.42, "tokens/trainable": 87265} +{"epoch": 0.75, "grad_norm": 0.24030189216136932, "learning_rate": 7.657199467573129e-05, "loss": 0.004537718836218119, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00455, "step": 192, "tokens/total": 5803136, "tokens/train_per_sec_per_gpu": 34.5, "tokens/trainable": 87747} +{"epoch": 0.75390625, "grad_norm": 0.17005977034568787, "learning_rate": 7.631678576708561e-05, "loss": 0.003625418059527874, "memory/device_reserved (GiB)": 38.58, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00363, "step": 193, "tokens/total": 5833440, "tokens/train_per_sec_per_gpu": 36.39, "tokens/trainable": 88239} +{"epoch": 0.7578125, "grad_norm": 0.22860349714756012, "learning_rate": 7.606068977997255e-05, "loss": 0.0007644169381819665, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00076, "step": 194, "tokens/total": 5863552, "tokens/train_per_sec_per_gpu": 36.27, "tokens/trainable": 88718} +{"epoch": 0.76171875, "grad_norm": 0.5368777513504028, "learning_rate": 7.580371737159148e-05, "loss": 0.015955395996570587, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01608, "step": 195, "tokens/total": 5893680, "tokens/train_per_sec_per_gpu": 31.42, "tokens/trainable": 89129} +{"epoch": 0.765625, "grad_norm": 0.04851720482110977, "learning_rate": 7.554587923561324e-05, "loss": 0.0004464306985028088, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00045, "step": 196, "tokens/total": 5923744, "tokens/train_per_sec_per_gpu": 32.58, "tokens/trainable": 89567} +{"epoch": 0.76953125, "grad_norm": 2.685831308364868, "learning_rate": 7.528718610173511e-05, "loss": 0.03162113204598427, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.03213, "step": 197, "tokens/total": 5954128, "tokens/train_per_sec_per_gpu": 30.71, "tokens/trainable": 89988} +{"epoch": 0.7734375, "grad_norm": 0.11396117508411407, "learning_rate": 7.502764873523431e-05, "loss": 0.0007245016167871654, "memory/device_reserved (GiB)": 35.56, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00072, "step": 198, "tokens/total": 5984352, "tokens/train_per_sec_per_gpu": 36.54, "tokens/trainable": 90434} +{"epoch": 0.77734375, "grad_norm": 0.4947444498538971, "learning_rate": 7.476727793652011e-05, "loss": 0.01100912131369114, "memory/device_reserved (GiB)": 35.6, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01107, "step": 199, "tokens/total": 6014704, "tokens/train_per_sec_per_gpu": 34.51, "tokens/trainable": 90913} +{"epoch": 0.78125, "grad_norm": 0.6466076374053955, "learning_rate": 7.450608454068415e-05, "loss": 0.01665530353784561, "memory/device_reserved (GiB)": 35.66, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.01679, "step": 200, "tokens/total": 6045056, "tokens/train_per_sec_per_gpu": 34.59, "tokens/trainable": 91355} +{"epoch": 0.78515625, "grad_norm": 0.33924102783203125, "learning_rate": 7.424407941704987e-05, "loss": 0.008527595549821854, "memory/device_reserved (GiB)": 35.66, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00856, "step": 201, "tokens/total": 6075504, "tokens/train_per_sec_per_gpu": 37.73, "tokens/trainable": 91856} +{"epoch": 0.7890625, "grad_norm": 0.11541225016117096, "learning_rate": 7.398127346871986e-05, "loss": 0.001006019883789122, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00101, "step": 202, "tokens/total": 6105904, "tokens/train_per_sec_per_gpu": 34.48, "tokens/trainable": 92311} +{"epoch": 0.79296875, "grad_norm": 0.013573888689279556, "learning_rate": 7.371767763212238e-05, "loss": 0.00013932021101936698, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00014, "step": 203, "tokens/total": 6136272, "tokens/train_per_sec_per_gpu": 35.02, "tokens/trainable": 92764} +{"epoch": 0.796875, "grad_norm": 0.3160369098186493, "learning_rate": 7.345330287655617e-05, "loss": 0.013633212074637413, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.65, "memory/max_allocated (GiB)": 33.65, "ppl": 1.01373, "step": 204, "tokens/total": 6166336, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 93227} +{"epoch": 0.80078125, "grad_norm": 0.2614762783050537, "learning_rate": 7.31881602037339e-05, "loss": 0.006595024839043617, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00662, "step": 205, "tokens/total": 6196720, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 93675} +{"epoch": 0.8046875, "grad_norm": 0.2725606858730316, "learning_rate": 7.29222606473245e-05, "loss": 0.005595955532044172, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00561, "step": 206, "tokens/total": 6227424, "tokens/train_per_sec_per_gpu": 32.68, "tokens/trainable": 94120} +{"epoch": 0.80859375, "grad_norm": 0.05208409205079079, "learning_rate": 7.265561527249383e-05, "loss": 0.0006884766044095159, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00069, "step": 207, "tokens/total": 6257616, "tokens/train_per_sec_per_gpu": 35.68, "tokens/trainable": 94557} +{"epoch": 0.8125, "grad_norm": 0.6857002973556519, "learning_rate": 7.238823517544436e-05, "loss": 0.010781912133097649, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01084, "step": 208, "tokens/total": 6288000, "tokens/train_per_sec_per_gpu": 40.97, "tokens/trainable": 95035} +{"epoch": 0.81640625, "grad_norm": 0.18070466816425323, "learning_rate": 7.212013148295333e-05, "loss": 0.002757919952273369, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00276, "step": 209, "tokens/total": 6318336, "tokens/train_per_sec_per_gpu": 37.87, "tokens/trainable": 95517} +{"epoch": 0.8203125, "grad_norm": 0.16365700960159302, "learning_rate": 7.185131535190975e-05, "loss": 0.005094663705676794, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00511, "step": 210, "tokens/total": 6348656, "tokens/train_per_sec_per_gpu": 36.74, "tokens/trainable": 96026} +{"epoch": 0.82421875, "grad_norm": 0.2167346179485321, "learning_rate": 7.158179796885005e-05, "loss": 0.007257625460624695, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00728, "step": 211, "tokens/total": 6379040, "tokens/train_per_sec_per_gpu": 35.83, "tokens/trainable": 96477} +{"epoch": 0.828125, "grad_norm": 0.08415517956018448, "learning_rate": 7.131159054949273e-05, "loss": 0.0018573766574263573, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00186, "step": 212, "tokens/total": 6409312, "tokens/train_per_sec_per_gpu": 33.39, "tokens/trainable": 96951} +{"epoch": 0.83203125, "grad_norm": 0.4260450601577759, "learning_rate": 7.104070433827139e-05, "loss": 0.021458052098751068, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.02169, "step": 213, "tokens/total": 6439520, "tokens/train_per_sec_per_gpu": 30.91, "tokens/trainable": 97350} +{"epoch": 0.8359375, "grad_norm": 0.13258714973926544, "learning_rate": 7.076915060786705e-05, "loss": 0.0031774642411619425, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00318, "step": 214, "tokens/total": 6469888, "tokens/train_per_sec_per_gpu": 30.04, "tokens/trainable": 97792} +{"epoch": 0.83984375, "grad_norm": 0.4242945909500122, "learning_rate": 7.049694065873882e-05, "loss": 0.017319604754447937, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01747, "step": 215, "tokens/total": 6500128, "tokens/train_per_sec_per_gpu": 36.65, "tokens/trainable": 98257} +{"epoch": 0.84375, "grad_norm": 0.3156379461288452, "learning_rate": 7.022408581865382e-05, "loss": 0.015216855332255363, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.01533, "step": 216, "tokens/total": 6530832, "tokens/train_per_sec_per_gpu": 32.87, "tokens/trainable": 98731} +{"epoch": 0.84765625, "grad_norm": 0.05867718160152435, "learning_rate": 6.99505974422157e-05, "loss": 0.00046426919288933277, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00046, "step": 217, "tokens/total": 6561248, "tokens/train_per_sec_per_gpu": 36.71, "tokens/trainable": 99212} +{"epoch": 0.8515625, "grad_norm": 0.0972680002450943, "learning_rate": 6.967648691039213e-05, "loss": 0.0036130475345999002, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00362, "step": 218, "tokens/total": 6591648, "tokens/train_per_sec_per_gpu": 33.27, "tokens/trainable": 99654} +{"epoch": 0.85546875, "grad_norm": 0.10315816849470139, "learning_rate": 6.940176563004123e-05, "loss": 0.0020681945607066154, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00207, "step": 219, "tokens/total": 6622368, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 100086} +{"epoch": 0.859375, "grad_norm": 0.09276516735553741, "learning_rate": 6.912644503343682e-05, "loss": 0.0020987873431295156, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0021, "step": 220, "tokens/total": 6652848, "tokens/train_per_sec_per_gpu": 33.54, "tokens/trainable": 100524} +{"epoch": 0.86328125, "grad_norm": 0.3432714343070984, "learning_rate": 6.885053657779273e-05, "loss": 0.006101437844336033, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00612, "step": 221, "tokens/total": 6683392, "tokens/train_per_sec_per_gpu": 33.63, "tokens/trainable": 100996} +{"epoch": 0.8671875, "grad_norm": 0.32484498620033264, "learning_rate": 6.857405174478604e-05, "loss": 0.005491882096976042, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00551, "step": 222, "tokens/total": 6713712, "tokens/train_per_sec_per_gpu": 32.22, "tokens/trainable": 101449} +{"epoch": 0.87109375, "grad_norm": 0.408407986164093, "learning_rate": 6.82970020400792e-05, "loss": 0.014964321628212929, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01508, "step": 223, "tokens/total": 6744176, "tokens/train_per_sec_per_gpu": 33.1, "tokens/trainable": 101905} +{"epoch": 0.875, "grad_norm": 0.07530955970287323, "learning_rate": 6.801939899284132e-05, "loss": 0.002201129449531436, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0022, "step": 224, "tokens/total": 6774416, "tokens/train_per_sec_per_gpu": 36.68, "tokens/trainable": 102411} +{"epoch": 0.87890625, "grad_norm": 0.07113490253686905, "learning_rate": 6.774125415526827e-05, "loss": 0.001676997751928866, "memory/device_reserved (GiB)": 37.19, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00168, "step": 225, "tokens/total": 6804592, "tokens/train_per_sec_per_gpu": 34.17, "tokens/trainable": 102882} +{"epoch": 0.8828125, "grad_norm": 0.02237725257873535, "learning_rate": 6.746257910210214e-05, "loss": 0.0004955717595294118, "memory/device_reserved (GiB)": 35.49, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0005, "step": 226, "tokens/total": 6834976, "tokens/train_per_sec_per_gpu": 34.85, "tokens/trainable": 103332} +{"epoch": 0.88671875, "grad_norm": 0.2688956558704376, "learning_rate": 6.718338543014937e-05, "loss": 0.01203202549368143, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0121, "step": 227, "tokens/total": 6865408, "tokens/train_per_sec_per_gpu": 35.35, "tokens/trainable": 103790} +{"epoch": 0.890625, "grad_norm": 0.11738378554582596, "learning_rate": 6.69036847577983e-05, "loss": 0.0008421347592957318, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00084, "step": 228, "tokens/total": 6895664, "tokens/train_per_sec_per_gpu": 34.56, "tokens/trainable": 104246} +{"epoch": 0.89453125, "grad_norm": 0.7972850203514099, "learning_rate": 6.662348872453553e-05, "loss": 0.006434948183596134, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00646, "step": 229, "tokens/total": 6926016, "tokens/train_per_sec_per_gpu": 37.92, "tokens/trainable": 104707} +{"epoch": 0.8984375, "grad_norm": 0.037981245666742325, "learning_rate": 6.63428089904618e-05, "loss": 0.0003998051688540727, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0004, "step": 230, "tokens/total": 6956384, "tokens/train_per_sec_per_gpu": 37.08, "tokens/trainable": 105167} +{"epoch": 0.90234375, "grad_norm": 0.06838542222976685, "learning_rate": 6.60616572358065e-05, "loss": 0.0011449911398813128, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00115, "step": 231, "tokens/total": 6986768, "tokens/train_per_sec_per_gpu": 33.11, "tokens/trainable": 105576} +{"epoch": 0.90625, "grad_norm": 0.0395381897687912, "learning_rate": 6.578004516044172e-05, "loss": 0.0007394620915874839, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00074, "step": 232, "tokens/total": 7017232, "tokens/train_per_sec_per_gpu": 36.98, "tokens/trainable": 106043} +{"epoch": 0.91015625, "grad_norm": 0.40814921259880066, "learning_rate": 6.549798448339548e-05, "loss": 0.005036160349845886, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00505, "step": 233, "tokens/total": 7047568, "tokens/train_per_sec_per_gpu": 35.52, "tokens/trainable": 106506} +{"epoch": 0.9140625, "grad_norm": 0.03264355659484863, "learning_rate": 6.521548694236384e-05, "loss": 0.0003136758750770241, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00031, "step": 234, "tokens/total": 7077760, "tokens/train_per_sec_per_gpu": 34.49, "tokens/trainable": 106951} +{"epoch": 0.91796875, "grad_norm": 0.06884738057851791, "learning_rate": 6.493256429322259e-05, "loss": 0.000555745791643858, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00056, "step": 235, "tokens/total": 7107760, "tokens/train_per_sec_per_gpu": 31.02, "tokens/trainable": 107382} +{"epoch": 0.921875, "grad_norm": 0.1841747760772705, "learning_rate": 6.464922830953799e-05, "loss": 0.003819565288722515, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00383, "step": 236, "tokens/total": 7138144, "tokens/train_per_sec_per_gpu": 33.18, "tokens/trainable": 107814} +{"epoch": 0.92578125, "grad_norm": 1.5293514728546143, "learning_rate": 6.436549078207688e-05, "loss": 0.02590152993798256, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.02624, "step": 237, "tokens/total": 7168352, "tokens/train_per_sec_per_gpu": 35.58, "tokens/trainable": 108274} +{"epoch": 0.9296875, "grad_norm": 0.06575962156057358, "learning_rate": 6.408136351831592e-05, "loss": 0.0009253169409930706, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00093, "step": 238, "tokens/total": 7198624, "tokens/train_per_sec_per_gpu": 29.35, "tokens/trainable": 108714} +{"epoch": 0.93359375, "grad_norm": 0.048577819019556046, "learning_rate": 6.379685834195036e-05, "loss": 0.0002521112619433552, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00025, "step": 239, "tokens/total": 7229216, "tokens/train_per_sec_per_gpu": 36.92, "tokens/trainable": 109220} +{"epoch": 0.9375, "grad_norm": 0.28133440017700195, "learning_rate": 6.351198709240186e-05, "loss": 0.003380296751856804, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00339, "step": 240, "tokens/total": 7259648, "tokens/train_per_sec_per_gpu": 32.28, "tokens/trainable": 109672} +{"epoch": 0.94140625, "grad_norm": 0.30106109380722046, "learning_rate": 6.32267616243259e-05, "loss": 0.0028490356635302305, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00285, "step": 241, "tokens/total": 7289824, "tokens/train_per_sec_per_gpu": 35.42, "tokens/trainable": 110135} +{"epoch": 0.9453125, "grad_norm": 0.4230552911758423, "learning_rate": 6.294119380711849e-05, "loss": 0.004133709240704775, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00414, "step": 242, "tokens/total": 7320288, "tokens/train_per_sec_per_gpu": 30.81, "tokens/trainable": 110563} +{"epoch": 0.94921875, "grad_norm": 0.2737712860107422, "learning_rate": 6.265529552442209e-05, "loss": 0.0038032641168683767, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00381, "step": 243, "tokens/total": 7350736, "tokens/train_per_sec_per_gpu": 37.33, "tokens/trainable": 111049} +{"epoch": 0.953125, "grad_norm": 0.16202203929424286, "learning_rate": 6.236907867363127e-05, "loss": 0.0011738601606339216, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00117, "step": 244, "tokens/total": 7381280, "tokens/train_per_sec_per_gpu": 33.62, "tokens/trainable": 111495} +{"epoch": 0.95703125, "grad_norm": 0.4818413555622101, "learning_rate": 6.208255516539749e-05, "loss": 0.011270345188677311, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01133, "step": 245, "tokens/total": 7411632, "tokens/train_per_sec_per_gpu": 32.03, "tokens/trainable": 111968} +{"epoch": 0.9609375, "grad_norm": 0.09653587639331818, "learning_rate": 6.179573692313344e-05, "loss": 0.0007583714323118329, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00076, "step": 246, "tokens/total": 7441904, "tokens/train_per_sec_per_gpu": 36.76, "tokens/trainable": 112416} +{"epoch": 0.96484375, "grad_norm": 0.07513346523046494, "learning_rate": 6.150863588251694e-05, "loss": 0.0008964500157162547, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0009, "step": 247, "tokens/total": 7472368, "tokens/train_per_sec_per_gpu": 34.22, "tokens/trainable": 112882} +{"epoch": 0.96875, "grad_norm": 0.069392628967762, "learning_rate": 6.122126399099419e-05, "loss": 0.001440772321075201, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00144, "step": 248, "tokens/total": 7502656, "tokens/train_per_sec_per_gpu": 40.27, "tokens/trainable": 113390} +{"epoch": 0.97265625, "grad_norm": 0.48798877000808716, "learning_rate": 6.0933633207282615e-05, "loss": 0.010719450190663338, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01078, "step": 249, "tokens/total": 7532800, "tokens/train_per_sec_per_gpu": 38.0, "tokens/trainable": 113904} +{"epoch": 0.9765625, "grad_norm": 0.2219225913286209, "learning_rate": 6.064575550087316e-05, "loss": 0.0012265645200386643, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00123, "step": 250, "tokens/total": 7563264, "tokens/train_per_sec_per_gpu": 34.14, "tokens/trainable": 114351} +{"epoch": 0.98046875, "grad_norm": 2.7423243522644043, "learning_rate": 6.0357642851532245e-05, "loss": 0.004063127096742392, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00407, "step": 251, "tokens/total": 7593840, "tokens/train_per_sec_per_gpu": 35.54, "tokens/trainable": 114842} +{"epoch": 0.984375, "grad_norm": 0.3986411690711975, "learning_rate": 6.0069307248803294e-05, "loss": 0.004339811857789755, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00435, "step": 252, "tokens/total": 7624416, "tokens/train_per_sec_per_gpu": 29.1, "tokens/trainable": 115263} +{"epoch": 0.98828125, "grad_norm": 0.23857781291007996, "learning_rate": 5.9780760691507635e-05, "loss": 0.003166551236063242, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00317, "step": 253, "tokens/total": 7654688, "tokens/train_per_sec_per_gpu": 33.0, "tokens/trainable": 115703} +{"epoch": 0.9921875, "grad_norm": 0.13943128287792206, "learning_rate": 5.9492015187245334e-05, "loss": 0.002166214631870389, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00217, "step": 254, "tokens/total": 7685088, "tokens/train_per_sec_per_gpu": 33.52, "tokens/trainable": 116157} +{"epoch": 0.99609375, "grad_norm": 0.5217611193656921, "learning_rate": 5.920308275189541e-05, "loss": 0.011248281225562096, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01131, "step": 255, "tokens/total": 7715552, "tokens/train_per_sec_per_gpu": 32.28, "tokens/trainable": 116567} +{"epoch": 1.0, "grad_norm": 0.12602734565734863, "learning_rate": 5.8913975409115874e-05, "loss": 0.002226680051535368, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00223, "step": 256, "tokens/total": 7745872, "tokens/train_per_sec_per_gpu": 31.48, "tokens/trainable": 117030} +{"epoch": 1.00390625, "grad_norm": 0.06412868201732635, "learning_rate": 5.8624705189843395e-05, "loss": 0.0004693535447586328, "memory/device_reserved (GiB)": 38.23, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00047, "step": 257, "tokens/total": 7776208, "tokens/train_per_sec_per_gpu": 34.26, "tokens/trainable": 117517} +{"epoch": 1.0078125, "grad_norm": 0.21263156831264496, "learning_rate": 5.833528413179249e-05, "loss": 0.0017973874928429723, "memory/device_reserved (GiB)": 35.62, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0018, "step": 258, "tokens/total": 7806480, "tokens/train_per_sec_per_gpu": 38.62, "tokens/trainable": 118015} +{"epoch": 1.01171875, "grad_norm": 0.16070301830768585, "learning_rate": 5.80457242789548e-05, "loss": 0.0026787673123180866, "memory/device_reserved (GiB)": 35.62, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00268, "step": 259, "tokens/total": 7836800, "tokens/train_per_sec_per_gpu": 36.58, "tokens/trainable": 118520} +{"epoch": 1.015625, "grad_norm": 0.09660778194665909, "learning_rate": 5.77560376810977e-05, "loss": 0.0018036369001492858, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00181, "step": 260, "tokens/total": 7867040, "tokens/train_per_sec_per_gpu": 36.34, "tokens/trainable": 118992} +{"epoch": 1.01953125, "grad_norm": 0.4718175232410431, "learning_rate": 5.7466236393263005e-05, "loss": 0.007852890528738499, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00788, "step": 261, "tokens/total": 7897408, "tokens/train_per_sec_per_gpu": 34.06, "tokens/trainable": 119438} +{"epoch": 1.0234375, "grad_norm": 0.2906242311000824, "learning_rate": 5.717633247526522e-05, "loss": 0.0019686920568346977, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00197, "step": 262, "tokens/total": 7927648, "tokens/train_per_sec_per_gpu": 32.53, "tokens/trainable": 119881} +{"epoch": 1.02734375, "grad_norm": 0.01843550242483616, "learning_rate": 5.688633799118971e-05, "loss": 0.0003263683756813407, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00033, "step": 263, "tokens/total": 7957968, "tokens/train_per_sec_per_gpu": 28.15, "tokens/trainable": 120285} +{"epoch": 1.03125, "grad_norm": 0.16429829597473145, "learning_rate": 5.659626500889066e-05, "loss": 0.0017057846998795867, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.65, "memory/max_allocated (GiB)": 33.65, "ppl": 1.00171, "step": 264, "tokens/total": 7987888, "tokens/train_per_sec_per_gpu": 33.51, "tokens/trainable": 120755} +{"epoch": 1.03515625, "grad_norm": 0.0684906542301178, "learning_rate": 5.6306125599488905e-05, "loss": 0.0006418414413928986, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00064, "step": 265, "tokens/total": 8018048, "tokens/train_per_sec_per_gpu": 35.94, "tokens/trainable": 121194} +{"epoch": 1.0390625, "grad_norm": 0.4636438488960266, "learning_rate": 5.601593183686955e-05, "loss": 0.007831420749425888, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00786, "step": 266, "tokens/total": 8048448, "tokens/train_per_sec_per_gpu": 37.0, "tokens/trainable": 121670} +{"epoch": 1.04296875, "grad_norm": 0.2955789566040039, "learning_rate": 5.572569579717961e-05, "loss": 0.009791678749024868, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00984, "step": 267, "tokens/total": 8078848, "tokens/train_per_sec_per_gpu": 34.02, "tokens/trainable": 122142} +{"epoch": 1.046875, "grad_norm": 0.31020334362983704, "learning_rate": 5.543542955832538e-05, "loss": 0.004724925849586725, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00474, "step": 268, "tokens/total": 8109360, "tokens/train_per_sec_per_gpu": 32.73, "tokens/trainable": 122585} +{"epoch": 1.05078125, "grad_norm": 0.08568736165761948, "learning_rate": 5.514514519946986e-05, "loss": 0.0010347942588850856, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00104, "step": 269, "tokens/total": 8139888, "tokens/train_per_sec_per_gpu": 39.26, "tokens/trainable": 123091} +{"epoch": 1.0546875, "grad_norm": 0.09669848531484604, "learning_rate": 5.485485480053015e-05, "loss": 0.001207374152727425, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00121, "step": 270, "tokens/total": 8170480, "tokens/train_per_sec_per_gpu": 31.16, "tokens/trainable": 123505} +{"epoch": 1.05859375, "grad_norm": 0.19564303755760193, "learning_rate": 5.4564570441674645e-05, "loss": 0.0023664506152272224, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.41, "memory/max_allocated (GiB)": 33.41, "ppl": 1.00237, "step": 271, "tokens/total": 8198848, "tokens/train_per_sec_per_gpu": 35.83, "tokens/trainable": 123953} +{"epoch": 1.0625, "grad_norm": 0.37416645884513855, "learning_rate": 5.42743042028204e-05, "loss": 0.014112848788499832, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01421, "step": 272, "tokens/total": 8229200, "tokens/train_per_sec_per_gpu": 28.96, "tokens/trainable": 124400} +{"epoch": 1.06640625, "grad_norm": 0.49067050218582153, "learning_rate": 5.3984068163130464e-05, "loss": 0.021221842616796494, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02145, "step": 273, "tokens/total": 8259536, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 124870} +{"epoch": 1.0703125, "grad_norm": 0.241201713681221, "learning_rate": 5.369387440051111e-05, "loss": 0.006090693175792694, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00611, "step": 274, "tokens/total": 8289904, "tokens/train_per_sec_per_gpu": 32.44, "tokens/trainable": 125333} +{"epoch": 1.07421875, "grad_norm": 0.019017836079001427, "learning_rate": 5.340373499110935e-05, "loss": 0.0002696830197237432, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00027, "step": 275, "tokens/total": 8320176, "tokens/train_per_sec_per_gpu": 37.08, "tokens/trainable": 125834} +{"epoch": 1.078125, "grad_norm": 0.011502111330628395, "learning_rate": 5.3113662008810304e-05, "loss": 0.00019440105825196952, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00019, "step": 276, "tokens/total": 8350832, "tokens/train_per_sec_per_gpu": 38.45, "tokens/trainable": 126316} +{"epoch": 1.08203125, "grad_norm": 0.26989492774009705, "learning_rate": 5.282366752473479e-05, "loss": 0.006379636470228434, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.0064, "step": 277, "tokens/total": 8380976, "tokens/train_per_sec_per_gpu": 35.68, "tokens/trainable": 126798} +{"epoch": 1.0859375, "grad_norm": 0.8612332940101624, "learning_rate": 5.2533763606737005e-05, "loss": 0.01699192076921463, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.01714, "step": 278, "tokens/total": 8411072, "tokens/train_per_sec_per_gpu": 29.66, "tokens/trainable": 127223} +{"epoch": 1.08984375, "grad_norm": 0.32299262285232544, "learning_rate": 5.224396231890232e-05, "loss": 0.00695072254166007, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.00697, "step": 279, "tokens/total": 8440736, "tokens/train_per_sec_per_gpu": 36.28, "tokens/trainable": 127672} +{"epoch": 1.09375, "grad_norm": 0.12054527550935745, "learning_rate": 5.195427572104522e-05, "loss": 0.0043365005403757095, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00435, "step": 280, "tokens/total": 8470944, "tokens/train_per_sec_per_gpu": 32.49, "tokens/trainable": 128116} +{"epoch": 1.09765625, "grad_norm": 0.10208329558372498, "learning_rate": 5.166471586820751e-05, "loss": 0.001871119486168027, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00187, "step": 281, "tokens/total": 8501104, "tokens/train_per_sec_per_gpu": 33.16, "tokens/trainable": 128589} +{"epoch": 1.1015625, "grad_norm": 0.20265866816043854, "learning_rate": 5.1375294810156615e-05, "loss": 0.006930817384272814, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00695, "step": 282, "tokens/total": 8531696, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 129036} +{"epoch": 1.10546875, "grad_norm": 0.30612000823020935, "learning_rate": 5.1086024590884144e-05, "loss": 0.009662694297730923, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00971, "step": 283, "tokens/total": 8561936, "tokens/train_per_sec_per_gpu": 35.58, "tokens/trainable": 129485} +{"epoch": 1.109375, "grad_norm": 0.06664866954088211, "learning_rate": 5.079691724810461e-05, "loss": 0.0016388238873332739, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00164, "step": 284, "tokens/total": 8592192, "tokens/train_per_sec_per_gpu": 29.9, "tokens/trainable": 129920} +{"epoch": 1.11328125, "grad_norm": 0.38304245471954346, "learning_rate": 5.0507984812754684e-05, "loss": 0.002592553850263357, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0026, "step": 285, "tokens/total": 8622432, "tokens/train_per_sec_per_gpu": 36.49, "tokens/trainable": 130417} +{"epoch": 1.1171875, "grad_norm": 0.06439055502414703, "learning_rate": 5.021923930849237e-05, "loss": 0.0009901091689243913, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00099, "step": 286, "tokens/total": 8652768, "tokens/train_per_sec_per_gpu": 36.33, "tokens/trainable": 130903} +{"epoch": 1.12109375, "grad_norm": 0.14527085423469543, "learning_rate": 4.99306927511967e-05, "loss": 0.002050987211987376, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00205, "step": 287, "tokens/total": 8683296, "tokens/train_per_sec_per_gpu": 35.89, "tokens/trainable": 131343} +{"epoch": 1.125, "grad_norm": 0.14959585666656494, "learning_rate": 4.964235714846775e-05, "loss": 0.0029775684233754873, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00298, "step": 288, "tokens/total": 8713504, "tokens/train_per_sec_per_gpu": 32.55, "tokens/trainable": 131763} +{"epoch": 1.12890625, "grad_norm": 0.012836308218538761, "learning_rate": 4.9354244499126866e-05, "loss": 0.00020321214105933905, "memory/device_reserved (GiB)": 35.98, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0002, "step": 289, "tokens/total": 8743888, "tokens/train_per_sec_per_gpu": 34.28, "tokens/trainable": 132229} +{"epoch": 1.1328125, "grad_norm": 0.01663857325911522, "learning_rate": 4.90663667927174e-05, "loss": 0.00027601540205068886, "memory/device_reserved (GiB)": 35.54, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00028, "step": 290, "tokens/total": 8774336, "tokens/train_per_sec_per_gpu": 35.72, "tokens/trainable": 132678} +{"epoch": 1.13671875, "grad_norm": 0.019089875742793083, "learning_rate": 4.877873600900581e-05, "loss": 0.0003022018354386091, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0003, "step": 291, "tokens/total": 8804816, "tokens/train_per_sec_per_gpu": 29.59, "tokens/trainable": 133136} +{"epoch": 1.140625, "grad_norm": 0.021187657490372658, "learning_rate": 4.849136411748306e-05, "loss": 0.00046957118320278823, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00047, "step": 292, "tokens/total": 8835024, "tokens/train_per_sec_per_gpu": 35.54, "tokens/trainable": 133608} +{"epoch": 1.14453125, "grad_norm": 0.21631699800491333, "learning_rate": 4.8204263076866574e-05, "loss": 0.005231338553130627, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00525, "step": 293, "tokens/total": 8865408, "tokens/train_per_sec_per_gpu": 40.74, "tokens/trainable": 134108} +{"epoch": 1.1484375, "grad_norm": 0.06148816645145416, "learning_rate": 4.791744483460251e-05, "loss": 0.0008393025491386652, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00084, "step": 294, "tokens/total": 8895824, "tokens/train_per_sec_per_gpu": 30.96, "tokens/trainable": 134541} +{"epoch": 1.15234375, "grad_norm": 0.05927408114075661, "learning_rate": 4.7630921326368736e-05, "loss": 0.0008574716048315167, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00086, "step": 295, "tokens/total": 8926336, "tokens/train_per_sec_per_gpu": 29.67, "tokens/trainable": 134966} +{"epoch": 1.15625, "grad_norm": 0.019883394241333008, "learning_rate": 4.7344704475577916e-05, "loss": 0.0003959762107115239, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0004, "step": 296, "tokens/total": 8956656, "tokens/train_per_sec_per_gpu": 33.77, "tokens/trainable": 135402} +{"epoch": 1.16015625, "grad_norm": 0.14877387881278992, "learning_rate": 4.705880619288153e-05, "loss": 0.002186706755310297, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00219, "step": 297, "tokens/total": 8986752, "tokens/train_per_sec_per_gpu": 30.22, "tokens/trainable": 135804} +{"epoch": 1.1640625, "grad_norm": 1.7873051166534424, "learning_rate": 4.677323837567412e-05, "loss": 0.004662086255848408, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00467, "step": 298, "tokens/total": 9017024, "tokens/train_per_sec_per_gpu": 31.34, "tokens/trainable": 136231} +{"epoch": 1.16796875, "grad_norm": 0.015957769006490707, "learning_rate": 4.6488012907598146e-05, "loss": 0.00035542104160413146, "memory/device_reserved (GiB)": 35.55, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00036, "step": 299, "tokens/total": 9047424, "tokens/train_per_sec_per_gpu": 34.5, "tokens/trainable": 136705} +{"epoch": 1.171875, "grad_norm": 0.21242187917232513, "learning_rate": 4.620314165804964e-05, "loss": 0.002182526746764779, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00218, "step": 300, "tokens/total": 9077648, "tokens/train_per_sec_per_gpu": 35.03, "tokens/trainable": 137178} +{"epoch": 1.17578125, "grad_norm": 0.2956356704235077, "learning_rate": 4.591863648168407e-05, "loss": 0.007447251118719578, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00748, "step": 301, "tokens/total": 9108032, "tokens/train_per_sec_per_gpu": 34.65, "tokens/trainable": 137643} +{"epoch": 1.1796875, "grad_norm": 0.19968345761299133, "learning_rate": 4.5634509217923135e-05, "loss": 0.002804415300488472, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00281, "step": 302, "tokens/total": 9138240, "tokens/train_per_sec_per_gpu": 32.17, "tokens/trainable": 138078} +{"epoch": 1.18359375, "grad_norm": 0.09324834495782852, "learning_rate": 4.535077169046201e-05, "loss": 0.001014861511066556, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00102, "step": 303, "tokens/total": 9168352, "tokens/train_per_sec_per_gpu": 35.4, "tokens/trainable": 138515} +{"epoch": 1.1875, "grad_norm": 0.18735024333000183, "learning_rate": 4.506743570677743e-05, "loss": 0.0008113987278193235, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00081, "step": 304, "tokens/total": 9198864, "tokens/train_per_sec_per_gpu": 35.45, "tokens/trainable": 138948} +{"epoch": 1.19140625, "grad_norm": 0.20801964402198792, "learning_rate": 4.478451305763618e-05, "loss": 0.0025226804427802563, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00253, "step": 305, "tokens/total": 9229104, "tokens/train_per_sec_per_gpu": 38.42, "tokens/trainable": 139408} +{"epoch": 1.1953125, "grad_norm": 0.12481324374675751, "learning_rate": 4.450201551660454e-05, "loss": 0.001147121423855424, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.38, "memory/max_allocated (GiB)": 33.38, "ppl": 1.00115, "step": 306, "tokens/total": 9257344, "tokens/train_per_sec_per_gpu": 33.71, "tokens/trainable": 139840} +{"epoch": 1.19921875, "grad_norm": 0.16472840309143066, "learning_rate": 4.4219954839558276e-05, "loss": 0.0030050487257540226, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00301, "step": 307, "tokens/total": 9287520, "tokens/train_per_sec_per_gpu": 33.2, "tokens/trainable": 140281} +{"epoch": 1.203125, "grad_norm": 0.018171297386288643, "learning_rate": 4.393834276419352e-05, "loss": 0.00021931837545707822, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00022, "step": 308, "tokens/total": 9317888, "tokens/train_per_sec_per_gpu": 37.74, "tokens/trainable": 140776} +{"epoch": 1.20703125, "grad_norm": 0.10386991500854492, "learning_rate": 4.36571910095382e-05, "loss": 0.0008659960003569722, "memory/device_reserved (GiB)": 35.72, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00087, "step": 309, "tokens/total": 9348256, "tokens/train_per_sec_per_gpu": 37.18, "tokens/trainable": 141228} +{"epoch": 1.2109375, "grad_norm": 0.020061908289790154, "learning_rate": 4.337651127546448e-05, "loss": 0.00014169455971568823, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00014, "step": 310, "tokens/total": 9378560, "tokens/train_per_sec_per_gpu": 37.7, "tokens/trainable": 141679} +{"epoch": 1.21484375, "grad_norm": 0.12189290672540665, "learning_rate": 4.3096315242201736e-05, "loss": 0.0012950387317687273, "memory/device_reserved (GiB)": 35.74, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.0013, "step": 311, "tokens/total": 9408848, "tokens/train_per_sec_per_gpu": 32.59, "tokens/trainable": 142122} +{"epoch": 1.21875, "grad_norm": 0.00976351834833622, "learning_rate": 4.2816614569850635e-05, "loss": 0.00013979707728140056, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00014, "step": 312, "tokens/total": 9439344, "tokens/train_per_sec_per_gpu": 33.36, "tokens/trainable": 142561} +{"epoch": 1.22265625, "grad_norm": 0.2909281253814697, "learning_rate": 4.2537420897897864e-05, "loss": 0.005748535506427288, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00577, "step": 313, "tokens/total": 9469792, "tokens/train_per_sec_per_gpu": 35.97, "tokens/trainable": 143046} +{"epoch": 1.2265625, "grad_norm": 0.03204086795449257, "learning_rate": 4.225874584473174e-05, "loss": 0.0002670975518412888, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00027, "step": 314, "tokens/total": 9500128, "tokens/train_per_sec_per_gpu": 33.26, "tokens/trainable": 143493} +{"epoch": 1.23046875, "grad_norm": 0.05219706520438194, "learning_rate": 4.19806010071587e-05, "loss": 0.00025036506121978164, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00025, "step": 315, "tokens/total": 9530480, "tokens/train_per_sec_per_gpu": 35.48, "tokens/trainable": 143956} +{"epoch": 1.234375, "grad_norm": 0.9549329876899719, "learning_rate": 4.170299795992081e-05, "loss": 0.01225995272397995, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01234, "step": 316, "tokens/total": 9560912, "tokens/train_per_sec_per_gpu": 34.18, "tokens/trainable": 144411} +{"epoch": 1.23828125, "grad_norm": 0.028625696897506714, "learning_rate": 4.142594825521398e-05, "loss": 0.00029421402723528445, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00029, "step": 317, "tokens/total": 9591344, "tokens/train_per_sec_per_gpu": 38.96, "tokens/trainable": 144892} +{"epoch": 1.2421875, "grad_norm": 0.01120083313435316, "learning_rate": 4.114946342220728e-05, "loss": 0.00013539142673835158, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00014, "step": 318, "tokens/total": 9621392, "tokens/train_per_sec_per_gpu": 32.93, "tokens/trainable": 145339} +{"epoch": 1.24609375, "grad_norm": 0.0018913366366177797, "learning_rate": 4.087355496656321e-05, "loss": 3.699818626046181e-05, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00004, "step": 319, "tokens/total": 9651728, "tokens/train_per_sec_per_gpu": 36.67, "tokens/trainable": 145811} +{"epoch": 1.25, "grad_norm": 0.24486027657985687, "learning_rate": 4.05982343699588e-05, "loss": 0.0035814084112644196, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00359, "step": 320, "tokens/total": 9682224, "tokens/train_per_sec_per_gpu": 39.04, "tokens/trainable": 146314} +{"epoch": 1.25390625, "grad_norm": 0.007224086206406355, "learning_rate": 4.0323513089607876e-05, "loss": 8.45451868372038e-05, "memory/device_reserved (GiB)": 37.95, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00008, "step": 321, "tokens/total": 9712656, "tokens/train_per_sec_per_gpu": 31.97, "tokens/trainable": 146781} +{"epoch": 1.2578125, "grad_norm": 0.10228509455919266, "learning_rate": 4.004940255778431e-05, "loss": 0.0019895497243851423, "memory/device_reserved (GiB)": 34.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00199, "step": 322, "tokens/total": 9743088, "tokens/train_per_sec_per_gpu": 35.81, "tokens/trainable": 147225} +{"epoch": 1.26171875, "grad_norm": 0.08815925568342209, "learning_rate": 3.977591418134619e-05, "loss": 0.0012562067713588476, "memory/device_reserved (GiB)": 35.17, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00126, "step": 323, "tokens/total": 9773808, "tokens/train_per_sec_per_gpu": 35.1, "tokens/trainable": 147673} +{"epoch": 1.265625, "grad_norm": 0.04117776080965996, "learning_rate": 3.95030593412612e-05, "loss": 0.00045941316056996584, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00046, "step": 324, "tokens/total": 9804016, "tokens/train_per_sec_per_gpu": 33.5, "tokens/trainable": 148103} +{"epoch": 1.26953125, "grad_norm": 0.10311771184206009, "learning_rate": 3.923084939213296e-05, "loss": 0.0007725170580670238, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00077, "step": 325, "tokens/total": 9834592, "tokens/train_per_sec_per_gpu": 31.94, "tokens/trainable": 148535} +{"epoch": 1.2734375, "grad_norm": 0.0661429911851883, "learning_rate": 3.895929566172861e-05, "loss": 0.001271560788154602, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00127, "step": 326, "tokens/total": 9864848, "tokens/train_per_sec_per_gpu": 34.38, "tokens/trainable": 148999} +{"epoch": 1.27734375, "grad_norm": 0.48627135157585144, "learning_rate": 3.868840945050728e-05, "loss": 0.008270882070064545, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00831, "step": 327, "tokens/total": 9895168, "tokens/train_per_sec_per_gpu": 31.92, "tokens/trainable": 149431} +{"epoch": 1.28125, "grad_norm": 0.12237530201673508, "learning_rate": 3.841820203114995e-05, "loss": 0.001470191404223442, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00147, "step": 328, "tokens/total": 9925696, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 149886} +{"epoch": 1.28515625, "grad_norm": 0.1192391961812973, "learning_rate": 3.814868464809027e-05, "loss": 0.001252716057933867, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00125, "step": 329, "tokens/total": 9955648, "tokens/train_per_sec_per_gpu": 31.31, "tokens/trainable": 150288} +{"epoch": 1.2890625, "grad_norm": 0.5366963744163513, "learning_rate": 3.787986851704667e-05, "loss": 0.0043250457383692265, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00433, "step": 330, "tokens/total": 9985856, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 150733} +{"epoch": 1.29296875, "grad_norm": 0.006924811284989119, "learning_rate": 3.7611764824555654e-05, "loss": 8.977689140010625e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00009, "step": 331, "tokens/total": 10016208, "tokens/train_per_sec_per_gpu": 37.12, "tokens/trainable": 151207} +{"epoch": 1.296875, "grad_norm": 0.050131287425756454, "learning_rate": 3.734438472750619e-05, "loss": 0.00035900447983294725, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00036, "step": 332, "tokens/total": 10046512, "tokens/train_per_sec_per_gpu": 37.97, "tokens/trainable": 151694} +{"epoch": 1.30078125, "grad_norm": 0.02984483353793621, "learning_rate": 3.707773935267552e-05, "loss": 0.0002764730015769601, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00028, "step": 333, "tokens/total": 10076944, "tokens/train_per_sec_per_gpu": 38.64, "tokens/trainable": 152188} +{"epoch": 1.3046875, "grad_norm": 0.004975741263478994, "learning_rate": 3.68118397962661e-05, "loss": 7.946819096105173e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00008, "step": 334, "tokens/total": 10107296, "tokens/train_per_sec_per_gpu": 30.9, "tokens/trainable": 152596} +{"epoch": 1.30859375, "grad_norm": 0.004153846297413111, "learning_rate": 3.654669712344384e-05, "loss": 5.989645796944387e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00006, "step": 335, "tokens/total": 10137568, "tokens/train_per_sec_per_gpu": 36.44, "tokens/trainable": 153052} +{"epoch": 1.3125, "grad_norm": 0.005283186212182045, "learning_rate": 3.628232236787763e-05, "loss": 4.799446833203547e-05, "memory/device_reserved (GiB)": 37.17, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00005, "step": 336, "tokens/total": 10167952, "tokens/train_per_sec_per_gpu": 30.43, "tokens/trainable": 153491} +{"epoch": 1.31640625, "grad_norm": 0.030474629253149033, "learning_rate": 3.6018726531280144e-05, "loss": 0.00027237640460953116, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00027, "step": 337, "tokens/total": 10198512, "tokens/train_per_sec_per_gpu": 40.59, "tokens/trainable": 153983} +{"epoch": 1.3203125, "grad_norm": 0.5984165072441101, "learning_rate": 3.575592058295017e-05, "loss": 0.009363370016217232, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00941, "step": 338, "tokens/total": 10228752, "tokens/train_per_sec_per_gpu": 37.0, "tokens/trainable": 154464} +{"epoch": 1.32421875, "grad_norm": 0.09123861789703369, "learning_rate": 3.549391545931585e-05, "loss": 0.0007300444412976503, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00073, "step": 339, "tokens/total": 10259104, "tokens/train_per_sec_per_gpu": 36.15, "tokens/trainable": 154924} +{"epoch": 1.328125, "grad_norm": 0.044113557785749435, "learning_rate": 3.5232722063479914e-05, "loss": 0.00047350398381240666, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00047, "step": 340, "tokens/total": 10289520, "tokens/train_per_sec_per_gpu": 31.05, "tokens/trainable": 155373} +{"epoch": 1.33203125, "grad_norm": 0.22140735387802124, "learning_rate": 3.49723512647657e-05, "loss": 0.0011342635843902826, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00113, "step": 341, "tokens/total": 10320016, "tokens/train_per_sec_per_gpu": 36.84, "tokens/trainable": 155873} +{"epoch": 1.3359375, "grad_norm": 0.12345008552074432, "learning_rate": 3.471281389826491e-05, "loss": 0.0013970056315883994, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0014, "step": 342, "tokens/total": 10350368, "tokens/train_per_sec_per_gpu": 30.7, "tokens/trainable": 156278} +{"epoch": 1.33984375, "grad_norm": 0.02416493557393551, "learning_rate": 3.4454120764386764e-05, "loss": 0.0003286268038209528, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00033, "step": 343, "tokens/total": 10380624, "tokens/train_per_sec_per_gpu": 35.73, "tokens/trainable": 156733} +{"epoch": 1.34375, "grad_norm": 0.26387253403663635, "learning_rate": 3.4196282628408526e-05, "loss": 0.001414710539393127, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00142, "step": 344, "tokens/total": 10411024, "tokens/train_per_sec_per_gpu": 34.17, "tokens/trainable": 157203} +{"epoch": 1.34765625, "grad_norm": 0.012582412920892239, "learning_rate": 3.3939310220027456e-05, "loss": 8.591653750045225e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00009, "step": 345, "tokens/total": 10441248, "tokens/train_per_sec_per_gpu": 38.09, "tokens/trainable": 157707} +{"epoch": 1.3515625, "grad_norm": 0.0038655376993119717, "learning_rate": 3.3683214232914404e-05, "loss": 5.268297536531463e-05, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00005, "step": 346, "tokens/total": 10471712, "tokens/train_per_sec_per_gpu": 35.24, "tokens/trainable": 158180} +{"epoch": 1.35546875, "grad_norm": 0.0408964529633522, "learning_rate": 3.342800532426873e-05, "loss": 0.00027673851582221687, "memory/device_reserved (GiB)": 38.13, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00028, "step": 347, "tokens/total": 10502288, "tokens/train_per_sec_per_gpu": 34.68, "tokens/trainable": 158646} +{"epoch": 1.359375, "grad_norm": 0.0886145755648613, "learning_rate": 3.317369411437484e-05, "loss": 0.0005747403483837843, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00057, "step": 348, "tokens/total": 10532304, "tokens/train_per_sec_per_gpu": 35.88, "tokens/trainable": 159115} +{"epoch": 1.36328125, "grad_norm": 0.11402814835309982, "learning_rate": 3.292029118616024e-05, "loss": 0.000894754019100219, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.0009, "step": 349, "tokens/total": 10562608, "tokens/train_per_sec_per_gpu": 31.71, "tokens/trainable": 159538} +{"epoch": 1.3671875, "grad_norm": 0.08461544662714005, "learning_rate": 3.266780708475511e-05, "loss": 0.00066529237665236, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00067, "step": 350, "tokens/total": 10592992, "tokens/train_per_sec_per_gpu": 33.45, "tokens/trainable": 160016} +{"epoch": 1.37109375, "grad_norm": 0.2829553782939911, "learning_rate": 3.241625231705354e-05, "loss": 0.0002811444574035704, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00028, "step": 351, "tokens/total": 10623328, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 160475} +{"epoch": 1.375, "grad_norm": 0.03589304909110069, "learning_rate": 3.216563735127618e-05, "loss": 0.00016819429583847523, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00017, "step": 352, "tokens/total": 10653632, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 160952} +{"epoch": 1.37890625, "grad_norm": 0.16046081483364105, "learning_rate": 3.191597261653475e-05, "loss": 0.0013975308975204825, "memory/device_reserved (GiB)": 38.4, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0014, "step": 353, "tokens/total": 10684176, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 161420} +{"epoch": 1.3828125, "grad_norm": 0.2950722277164459, "learning_rate": 3.166726850239794e-05, "loss": 0.00403338810428977, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00404, "step": 354, "tokens/total": 10714352, "tokens/train_per_sec_per_gpu": 34.38, "tokens/trainable": 161868} +{"epoch": 1.38671875, "grad_norm": 0.0004486274265218526, "learning_rate": 3.141953535845912e-05, "loss": 1.0177070180361625e-05, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00001, "step": 355, "tokens/total": 10744464, "tokens/train_per_sec_per_gpu": 35.73, "tokens/trainable": 162318} +{"epoch": 1.390625, "grad_norm": 0.01730065606534481, "learning_rate": 3.11727834939056e-05, "loss": 7.366175123024732e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00007, "step": 356, "tokens/total": 10774976, "tokens/train_per_sec_per_gpu": 31.71, "tokens/trainable": 162789} +{"epoch": 1.39453125, "grad_norm": 0.4219278395175934, "learning_rate": 3.092702317708967e-05, "loss": 0.006937004160135984, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00696, "step": 357, "tokens/total": 10805600, "tokens/train_per_sec_per_gpu": 36.51, "tokens/trainable": 163254} +{"epoch": 1.3984375, "grad_norm": 0.008418221957981586, "learning_rate": 3.0682264635101276e-05, "loss": 5.021447941544466e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00005, "step": 358, "tokens/total": 10835920, "tokens/train_per_sec_per_gpu": 34.74, "tokens/trainable": 163715} +{"epoch": 1.40234375, "grad_norm": 0.3869379460811615, "learning_rate": 3.0438518053342407e-05, "loss": 0.0042590186931192875, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00427, "step": 359, "tokens/total": 10866048, "tokens/train_per_sec_per_gpu": 34.73, "tokens/trainable": 164197} +{"epoch": 1.40625, "grad_norm": 0.2765217125415802, "learning_rate": 3.0195793575103266e-05, "loss": 0.0020075999200344086, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00201, "step": 360, "tokens/total": 10896288, "tokens/train_per_sec_per_gpu": 31.86, "tokens/trainable": 164661} +{"epoch": 1.41015625, "grad_norm": 0.2115616798400879, "learning_rate": 2.9954101301140146e-05, "loss": 0.0020617658738046885, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00206, "step": 361, "tokens/total": 10926816, "tokens/train_per_sec_per_gpu": 33.35, "tokens/trainable": 165111} +{"epoch": 1.4140625, "grad_norm": 0.011683800257742405, "learning_rate": 2.9713451289255123e-05, "loss": 0.0001429952826583758, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.31, "memory/max_allocated (GiB)": 33.31, "ppl": 1.00014, "step": 362, "tokens/total": 10954928, "tokens/train_per_sec_per_gpu": 30.67, "tokens/trainable": 165552} +{"epoch": 1.41796875, "grad_norm": 0.0025901086628437042, "learning_rate": 2.9473853553877484e-05, "loss": 2.3585678718518466e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00002, "step": 363, "tokens/total": 10985328, "tokens/train_per_sec_per_gpu": 35.54, "tokens/trainable": 166026} +{"epoch": 1.421875, "grad_norm": 0.14587508141994476, "learning_rate": 2.9235318065647e-05, "loss": 0.0011214152909815311, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00112, "step": 364, "tokens/total": 11015664, "tokens/train_per_sec_per_gpu": 38.23, "tokens/trainable": 166486} +{"epoch": 1.42578125, "grad_norm": 0.021550685167312622, "learning_rate": 2.8997854750998964e-05, "loss": 0.00018227534019388258, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00018, "step": 365, "tokens/total": 11046256, "tokens/train_per_sec_per_gpu": 36.71, "tokens/trainable": 166959} +{"epoch": 1.4296875, "grad_norm": 0.002367217792198062, "learning_rate": 2.8761473491751258e-05, "loss": 2.783137279038783e-05, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.68, "memory/max_allocated (GiB)": 33.68, "ppl": 1.00003, "step": 366, "tokens/total": 11076288, "tokens/train_per_sec_per_gpu": 33.67, "tokens/trainable": 167394} +{"epoch": 1.43359375, "grad_norm": 0.02959338203072548, "learning_rate": 2.8526184124692883e-05, "loss": 0.0001235282834386453, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00012, "step": 367, "tokens/total": 11106416, "tokens/train_per_sec_per_gpu": 28.63, "tokens/trainable": 167832} +{"epoch": 1.4375, "grad_norm": 0.04836918041110039, "learning_rate": 2.829199644117484e-05, "loss": 0.0002190673112636432, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00022, "step": 368, "tokens/total": 11136848, "tokens/train_per_sec_per_gpu": 31.23, "tokens/trainable": 168242} +{"epoch": 1.44140625, "grad_norm": 0.17935794591903687, "learning_rate": 2.8058920186702553e-05, "loss": 0.001091235550120473, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00109, "step": 369, "tokens/total": 11167376, "tokens/train_per_sec_per_gpu": 36.0, "tokens/trainable": 168715} +{"epoch": 1.4453125, "grad_norm": 0.17555589973926544, "learning_rate": 2.782696506053033e-05, "loss": 0.0008262722985818982, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00083, "step": 370, "tokens/total": 11197776, "tokens/train_per_sec_per_gpu": 32.06, "tokens/trainable": 169147} +{"epoch": 1.44921875, "grad_norm": 0.015604093670845032, "learning_rate": 2.7596140715257824e-05, "loss": 7.283307058969513e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00007, "step": 371, "tokens/total": 11227840, "tokens/train_per_sec_per_gpu": 34.57, "tokens/trainable": 169628} +{"epoch": 1.453125, "grad_norm": 0.001744552399031818, "learning_rate": 2.7366456756428184e-05, "loss": 1.2661729670071509e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00001, "step": 372, "tokens/total": 11258176, "tokens/train_per_sec_per_gpu": 35.48, "tokens/trainable": 170085} +{"epoch": 1.45703125, "grad_norm": 0.03369157761335373, "learning_rate": 2.7137922742128486e-05, "loss": 0.00013295823009684682, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00013, "step": 373, "tokens/total": 11288336, "tokens/train_per_sec_per_gpu": 35.9, "tokens/trainable": 170584} +{"epoch": 1.4609375, "grad_norm": 0.11042775958776474, "learning_rate": 2.691054818259188e-05, "loss": 0.00030668824911117554, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00031, "step": 374, "tokens/total": 11318848, "tokens/train_per_sec_per_gpu": 34.92, "tokens/trainable": 171058} +{"epoch": 1.46484375, "grad_norm": 0.015241619199514389, "learning_rate": 2.6684342539801933e-05, "loss": 5.481348853209056e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00005, "step": 375, "tokens/total": 11349168, "tokens/train_per_sec_per_gpu": 34.09, "tokens/trainable": 171518} +{"epoch": 1.46875, "grad_norm": 0.07663743942975998, "learning_rate": 2.645931522709877e-05, "loss": 0.0004268307238817215, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00043, "step": 376, "tokens/total": 11379328, "tokens/train_per_sec_per_gpu": 40.72, "tokens/trainable": 172007} +{"epoch": 1.47265625, "grad_norm": 0.3878970146179199, "learning_rate": 2.6235475608787365e-05, "loss": 0.006505412515252829, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00653, "step": 377, "tokens/total": 11409568, "tokens/train_per_sec_per_gpu": 34.98, "tokens/trainable": 172468} +{"epoch": 1.4765625, "grad_norm": 0.011625024490058422, "learning_rate": 2.6012832999747916e-05, "loss": 7.441644265782088e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00007, "step": 378, "tokens/total": 11439968, "tokens/train_per_sec_per_gpu": 33.97, "tokens/trainable": 172923} +{"epoch": 1.48046875, "grad_norm": 0.8276505470275879, "learning_rate": 2.579139666504821e-05, "loss": 0.022178810089826584, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.02243, "step": 379, "tokens/total": 11470496, "tokens/train_per_sec_per_gpu": 34.38, "tokens/trainable": 173370} +{"epoch": 1.484375, "grad_norm": 0.0010690620401874185, "learning_rate": 2.557117581955798e-05, "loss": 1.1797061233664863e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00001, "step": 380, "tokens/total": 11500720, "tokens/train_per_sec_per_gpu": 34.68, "tokens/trainable": 173837} +{"epoch": 1.48828125, "grad_norm": 0.06872054934501648, "learning_rate": 2.5352179627565532e-05, "loss": 0.0004525431140791625, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00045, "step": 381, "tokens/total": 11531280, "tokens/train_per_sec_per_gpu": 35.87, "tokens/trainable": 174294} +{"epoch": 1.4921875, "grad_norm": 0.025393052026629448, "learning_rate": 2.5134417202396277e-05, "loss": 0.0001451169518986717, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00015, "step": 382, "tokens/total": 11561264, "tokens/train_per_sec_per_gpu": 31.04, "tokens/trainable": 174701} +{"epoch": 1.49609375, "grad_norm": 0.006796801462769508, "learning_rate": 2.491789760603361e-05, "loss": 9.397654503118247e-05, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00009, "step": 383, "tokens/total": 11591616, "tokens/train_per_sec_per_gpu": 32.35, "tokens/trainable": 175131} +{"epoch": 1.5, "grad_norm": 0.35059845447540283, "learning_rate": 2.4702629848741764e-05, "loss": 0.004584255628287792, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00459, "step": 384, "tokens/total": 11622000, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 175586} +{"epoch": 1.50390625, "grad_norm": 0.33106186985969543, "learning_rate": 2.4488622888690785e-05, "loss": 0.010121386498212814, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01017, "step": 385, "tokens/total": 11652352, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 176026} +{"epoch": 1.5078125, "grad_norm": 0.003417555009946227, "learning_rate": 2.427588563158384e-05, "loss": 4.207541496725753e-05, "memory/device_reserved (GiB)": 34.98, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00004, "step": 386, "tokens/total": 11682736, "tokens/train_per_sec_per_gpu": 34.05, "tokens/trainable": 176512} +{"epoch": 1.51171875, "grad_norm": 0.24088126420974731, "learning_rate": 2.406442693028651e-05, "loss": 0.004325853195041418, "memory/device_reserved (GiB)": 34.98, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00434, "step": 387, "tokens/total": 11713136, "tokens/train_per_sec_per_gpu": 30.64, "tokens/trainable": 176943} +{"epoch": 1.515625, "grad_norm": 0.04677414149045944, "learning_rate": 2.3854255584458547e-05, "loss": 0.00039886703598313034, "memory/device_reserved (GiB)": 35.23, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.0004, "step": 388, "tokens/total": 11743808, "tokens/train_per_sec_per_gpu": 32.3, "tokens/trainable": 177370} +{"epoch": 1.51953125, "grad_norm": 0.01845633052289486, "learning_rate": 2.3645380340187508e-05, "loss": 0.00019935239106416702, "memory/device_reserved (GiB)": 35.58, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0002, "step": 389, "tokens/total": 11774144, "tokens/train_per_sec_per_gpu": 35.42, "tokens/trainable": 177851} +{"epoch": 1.5234375, "grad_norm": 0.19680730998516083, "learning_rate": 2.3437809889624914e-05, "loss": 0.0020088059827685356, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00201, "step": 390, "tokens/total": 11804272, "tokens/train_per_sec_per_gpu": 35.03, "tokens/trainable": 178310} +{"epoch": 1.52734375, "grad_norm": 0.029754316434264183, "learning_rate": 2.3231552870624487e-05, "loss": 0.00013066174869891256, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00013, "step": 391, "tokens/total": 11832432, "tokens/train_per_sec_per_gpu": 37.12, "tokens/trainable": 178736} +{"epoch": 1.53125, "grad_norm": 0.47083696722984314, "learning_rate": 2.3026617866382657e-05, "loss": 0.006020379718393087, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00604, "step": 392, "tokens/total": 11862960, "tokens/train_per_sec_per_gpu": 36.57, "tokens/trainable": 179233} +{"epoch": 1.53515625, "grad_norm": 0.02095848135650158, "learning_rate": 2.2823013405081507e-05, "loss": 0.00023951532784849405, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00024, "step": 393, "tokens/total": 11893232, "tokens/train_per_sec_per_gpu": 38.68, "tokens/trainable": 179730} +{"epoch": 1.5390625, "grad_norm": 0.09784477949142456, "learning_rate": 2.2620747959533722e-05, "loss": 0.0008653341792523861, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00087, "step": 394, "tokens/total": 11923664, "tokens/train_per_sec_per_gpu": 37.87, "tokens/trainable": 180227} +{"epoch": 1.54296875, "grad_norm": 0.016663849353790283, "learning_rate": 2.2419829946830123e-05, "loss": 0.0002501910785213113, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00025, "step": 395, "tokens/total": 11954032, "tokens/train_per_sec_per_gpu": 32.98, "tokens/trainable": 180687} +{"epoch": 1.546875, "grad_norm": 0.06284014135599136, "learning_rate": 2.2220267727989325e-05, "loss": 0.0006274969782680273, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00063, "step": 396, "tokens/total": 11984512, "tokens/train_per_sec_per_gpu": 33.05, "tokens/trainable": 181141} +{"epoch": 1.55078125, "grad_norm": 0.06682170182466507, "learning_rate": 2.202206960760984e-05, "loss": 0.0008653995464555919, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00087, "step": 397, "tokens/total": 12014880, "tokens/train_per_sec_per_gpu": 36.25, "tokens/trainable": 181648} +{"epoch": 1.5546875, "grad_norm": 0.048094261437654495, "learning_rate": 2.182524383352446e-05, "loss": 0.0006572521524503827, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00066, "step": 398, "tokens/total": 12044928, "tokens/train_per_sec_per_gpu": 34.17, "tokens/trainable": 182109} +{"epoch": 1.55859375, "grad_norm": 0.288680762052536, "learning_rate": 2.1629798596457056e-05, "loss": 0.0029174773953855038, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00292, "step": 399, "tokens/total": 12075344, "tokens/train_per_sec_per_gpu": 31.81, "tokens/trainable": 182559} +{"epoch": 1.5625, "grad_norm": 0.009714324958622456, "learning_rate": 2.1435742029681725e-05, "loss": 8.201875607483089e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00008, "step": 400, "tokens/total": 12105616, "tokens/train_per_sec_per_gpu": 33.54, "tokens/trainable": 182985} +{"epoch": 1.56640625, "grad_norm": 0.00939064659178257, "learning_rate": 2.124308220868431e-05, "loss": 0.00012157741002738476, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00012, "step": 401, "tokens/total": 12136080, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 183440} +{"epoch": 1.5703125, "grad_norm": 0.035272955894470215, "learning_rate": 2.105182715082638e-05, "loss": 0.0004544599214568734, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00045, "step": 402, "tokens/total": 12166240, "tokens/train_per_sec_per_gpu": 34.8, "tokens/trainable": 183917} +{"epoch": 1.57421875, "grad_norm": 0.22718806564807892, "learning_rate": 2.0861984815011552e-05, "loss": 0.0024650408886373043, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00247, "step": 403, "tokens/total": 12196544, "tokens/train_per_sec_per_gpu": 30.51, "tokens/trainable": 184348} +{"epoch": 1.578125, "grad_norm": 0.013060114346444607, "learning_rate": 2.0673563101354323e-05, "loss": 0.0001124740592786111, "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": 31.69, "tokens/trainable": 184786} +{"epoch": 1.58203125, "grad_norm": 0.052736442536115646, "learning_rate": 2.0486569850851317e-05, "loss": 0.0003598350449465215, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00036, "step": 405, "tokens/total": 12257264, "tokens/train_per_sec_per_gpu": 34.16, "tokens/trainable": 185249} +{"epoch": 1.5859375, "grad_norm": 0.00711337523534894, "learning_rate": 2.0301012845054956e-05, "loss": 8.437763608526438e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00008, "step": 406, "tokens/total": 12287584, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 185680} +{"epoch": 1.58984375, "grad_norm": 0.00903877429664135, "learning_rate": 2.011689980574966e-05, "loss": 0.00010423596540931612, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0001, "step": 407, "tokens/total": 12317984, "tokens/train_per_sec_per_gpu": 31.94, "tokens/trainable": 186100} +{"epoch": 1.59375, "grad_norm": 0.6410778164863586, "learning_rate": 1.993423839463052e-05, "loss": 0.004257278051227331, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00427, "step": 408, "tokens/total": 12348144, "tokens/train_per_sec_per_gpu": 40.16, "tokens/trainable": 186565} +{"epoch": 1.59765625, "grad_norm": 0.013556886464357376, "learning_rate": 1.975303621298445e-05, "loss": 0.0001645060838200152, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00016, "step": 409, "tokens/total": 12378544, "tokens/train_per_sec_per_gpu": 36.06, "tokens/trainable": 187017} +{"epoch": 1.6015625, "grad_norm": 0.10837272554636002, "learning_rate": 1.957330080137385e-05, "loss": 0.0001806107466109097, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00018, "step": 410, "tokens/total": 12408784, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 187468} +{"epoch": 1.60546875, "grad_norm": 0.005454800557345152, "learning_rate": 1.9395039639322864e-05, "loss": 9.649700950831175e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.0001, "step": 411, "tokens/total": 12438832, "tokens/train_per_sec_per_gpu": 37.94, "tokens/trainable": 187961} +{"epoch": 1.609375, "grad_norm": 0.003915195353329182, "learning_rate": 1.9218260145006073e-05, "loss": 5.392331149778329e-05, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00005, "step": 412, "tokens/total": 12469296, "tokens/train_per_sec_per_gpu": 36.08, "tokens/trainable": 188447} +{"epoch": 1.61328125, "grad_norm": 0.024293167516589165, "learning_rate": 1.904296967493982e-05, "loss": 0.00030749564757570624, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00031, "step": 413, "tokens/total": 12499392, "tokens/train_per_sec_per_gpu": 36.19, "tokens/trainable": 188903} +{"epoch": 1.6171875, "grad_norm": 0.1073823794722557, "learning_rate": 1.8869175523676064e-05, "loss": 0.0014971208292990923, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 34.0, "memory/max_allocated (GiB)": 34.0, "ppl": 1.0015, "step": 414, "tokens/total": 12529952, "tokens/train_per_sec_per_gpu": 36.87, "tokens/trainable": 189398} +{"epoch": 1.62109375, "grad_norm": 0.025999421253800392, "learning_rate": 1.869688492349885e-05, "loss": 0.00022489691036753356, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00022, "step": 415, "tokens/total": 12560240, "tokens/train_per_sec_per_gpu": 34.46, "tokens/trainable": 189850} +{"epoch": 1.625, "grad_norm": 0.033106278628110886, "learning_rate": 1.85261050441233e-05, "loss": 0.0003824468585662544, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00038, "step": 416, "tokens/total": 12590736, "tokens/train_per_sec_per_gpu": 32.95, "tokens/trainable": 190306} +{"epoch": 1.62890625, "grad_norm": 0.010407397523522377, "learning_rate": 1.8356842992397304e-05, "loss": 8.543908188585192e-05, "memory/device_reserved (GiB)": 38.05, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00009, "step": 417, "tokens/total": 12621168, "tokens/train_per_sec_per_gpu": 34.42, "tokens/trainable": 190746} +{"epoch": 1.6328125, "grad_norm": 0.007676505949348211, "learning_rate": 1.8189105812005714e-05, "loss": 9.762673289515078e-05, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.0001, "step": 418, "tokens/total": 12651456, "tokens/train_per_sec_per_gpu": 31.46, "tokens/trainable": 191156} +{"epoch": 1.63671875, "grad_norm": 0.008882309310138226, "learning_rate": 1.802290048317732e-05, "loss": 0.00011588008783292025, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00012, "step": 419, "tokens/total": 12681760, "tokens/train_per_sec_per_gpu": 29.78, "tokens/trainable": 191573} +{"epoch": 1.640625, "grad_norm": 0.01079952996224165, "learning_rate": 1.785823392239424e-05, "loss": 0.00015772700135130435, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00016, "step": 420, "tokens/total": 12712144, "tokens/train_per_sec_per_gpu": 33.14, "tokens/trainable": 192014} +{"epoch": 1.64453125, "grad_norm": 0.10464993119239807, "learning_rate": 1.7695112982104225e-05, "loss": 0.0007448438555002213, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00075, "step": 421, "tokens/total": 12742400, "tokens/train_per_sec_per_gpu": 33.16, "tokens/trainable": 192453} +{"epoch": 1.6484375, "grad_norm": 0.033524010330438614, "learning_rate": 1.7533544450435433e-05, "loss": 0.0003343333082739264, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00033, "step": 422, "tokens/total": 12772288, "tokens/train_per_sec_per_gpu": 33.53, "tokens/trainable": 192886} +{"epoch": 1.65234375, "grad_norm": 0.0037654207553714514, "learning_rate": 1.7373535050913946e-05, "loss": 4.5559077989310026e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00005, "step": 423, "tokens/total": 12802880, "tokens/train_per_sec_per_gpu": 30.19, "tokens/trainable": 193323} +{"epoch": 1.65625, "grad_norm": 0.06157264485955238, "learning_rate": 1.721509144218405e-05, "loss": 0.0005529897753149271, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00055, "step": 424, "tokens/total": 12833456, "tokens/train_per_sec_per_gpu": 32.84, "tokens/trainable": 193753} +{"epoch": 1.66015625, "grad_norm": 0.009139563888311386, "learning_rate": 1.705822021773101e-05, "loss": 0.00013727511395700276, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00014, "step": 425, "tokens/total": 12863952, "tokens/train_per_sec_per_gpu": 36.43, "tokens/trainable": 194240} +{"epoch": 1.6640625, "grad_norm": 0.32006195187568665, "learning_rate": 1.69029279056068e-05, "loss": 0.005133518483489752, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00515, "step": 426, "tokens/total": 12894352, "tokens/train_per_sec_per_gpu": 35.78, "tokens/trainable": 194680} +{"epoch": 1.66796875, "grad_norm": 0.19554191827774048, "learning_rate": 1.6749220968158415e-05, "loss": 0.0020202554296702147, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00202, "step": 427, "tokens/total": 12924688, "tokens/train_per_sec_per_gpu": 39.09, "tokens/trainable": 195186} +{"epoch": 1.671875, "grad_norm": 0.010584031231701374, "learning_rate": 1.659710580175893e-05, "loss": 7.440667104674503e-05, "memory/device_reserved (GiB)": 36.05, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00007, "step": 428, "tokens/total": 12955072, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 195693} +{"epoch": 1.67578125, "grad_norm": 0.005364356562495232, "learning_rate": 1.644658873654133e-05, "loss": 6.896692502778023e-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": 37.33, "tokens/trainable": 196130} +{"epoch": 1.6796875, "grad_norm": 0.0026337867602705956, "learning_rate": 1.629767603613508e-05, "loss": 5.40713153895922e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00005, "step": 430, "tokens/total": 13015600, "tokens/train_per_sec_per_gpu": 29.85, "tokens/trainable": 196530} +{"epoch": 1.68359375, "grad_norm": 0.08371307700872421, "learning_rate": 1.615037389740547e-05, "loss": 0.000647605920676142, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00065, "step": 431, "tokens/total": 13045728, "tokens/train_per_sec_per_gpu": 30.2, "tokens/trainable": 196974} +{"epoch": 1.6875, "grad_norm": 0.12106083333492279, "learning_rate": 1.600468845019576e-05, "loss": 0.0010971315205097198, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0011, "step": 432, "tokens/total": 13075920, "tokens/train_per_sec_per_gpu": 35.55, "tokens/trainable": 197454} +{"epoch": 1.69140625, "grad_norm": 0.0016024125507101417, "learning_rate": 1.5860625757072092e-05, "loss": 2.5408877263544127e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00003, "step": 433, "tokens/total": 13106128, "tokens/train_per_sec_per_gpu": 31.68, "tokens/trainable": 197902} +{"epoch": 1.6953125, "grad_norm": 0.006628716830164194, "learning_rate": 1.571819181307116e-05, "loss": 9.017687989398837e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00009, "step": 434, "tokens/total": 13136368, "tokens/train_per_sec_per_gpu": 36.0, "tokens/trainable": 198374} +{"epoch": 1.69921875, "grad_norm": 0.023524144664406776, "learning_rate": 1.557739254545075e-05, "loss": 0.0003021995653398335, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.0003, "step": 435, "tokens/total": 13166704, "tokens/train_per_sec_per_gpu": 37.86, "tokens/trainable": 198846} +{"epoch": 1.703125, "grad_norm": 0.1354794055223465, "learning_rate": 1.543823381344311e-05, "loss": 0.001777412835508585, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00178, "step": 436, "tokens/total": 13197344, "tokens/train_per_sec_per_gpu": 38.06, "tokens/trainable": 199320} +{"epoch": 1.70703125, "grad_norm": 0.0022158098872750998, "learning_rate": 1.5300721408011114e-05, "loss": 2.7318394131725654e-05, "memory/device_reserved (GiB)": 36.06, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00003, "step": 437, "tokens/total": 13227664, "tokens/train_per_sec_per_gpu": 37.77, "tokens/trainable": 199777} +{"epoch": 1.7109375, "grad_norm": 0.005856889300048351, "learning_rate": 1.5164861051607254e-05, "loss": 7.72368221078068e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00008, "step": 438, "tokens/total": 13257632, "tokens/train_per_sec_per_gpu": 31.16, "tokens/trainable": 200194} +{"epoch": 1.71484375, "grad_norm": 0.007335511967539787, "learning_rate": 1.5030658397935521e-05, "loss": 7.384424679912627e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00007, "step": 439, "tokens/total": 13288240, "tokens/train_per_sec_per_gpu": 34.14, "tokens/trainable": 200663} +{"epoch": 1.71875, "grad_norm": 0.00513842236250639, "learning_rate": 1.4898119031716104e-05, "loss": 8.217831782530993e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00008, "step": 440, "tokens/total": 13318992, "tokens/train_per_sec_per_gpu": 34.16, "tokens/trainable": 201121} +{"epoch": 1.72265625, "grad_norm": 0.0026071977335959673, "learning_rate": 1.476724846845306e-05, "loss": 3.909884981112555e-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": 40.13, "tokens/trainable": 201598} +{"epoch": 1.7265625, "grad_norm": 0.0043723853304982185, "learning_rate": 1.463805215420471e-05, "loss": 4.3211588490521535e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00004, "step": 442, "tokens/total": 13379648, "tokens/train_per_sec_per_gpu": 35.37, "tokens/trainable": 202068} +{"epoch": 1.73046875, "grad_norm": 0.032961804419755936, "learning_rate": 1.451053546535705e-05, "loss": 0.00024515652330592275, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00025, "step": 443, "tokens/total": 13409936, "tokens/train_per_sec_per_gpu": 35.82, "tokens/trainable": 202524} +{"epoch": 1.734375, "grad_norm": 0.12496478110551834, "learning_rate": 1.438470370840001e-05, "loss": 0.0008977800025604665, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.0009, "step": 444, "tokens/total": 13440288, "tokens/train_per_sec_per_gpu": 36.64, "tokens/trainable": 202996} +{"epoch": 1.73828125, "grad_norm": 0.016452984884381294, "learning_rate": 1.4260562119706606e-05, "loss": 0.00011743961658794433, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00012, "step": 445, "tokens/total": 13468880, "tokens/train_per_sec_per_gpu": 39.58, "tokens/trainable": 203470} +{"epoch": 1.7421875, "grad_norm": 0.009045973420143127, "learning_rate": 1.413811586531508e-05, "loss": 8.700705802766606e-05, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00009, "step": 446, "tokens/total": 13499184, "tokens/train_per_sec_per_gpu": 35.6, "tokens/trainable": 203919} +{"epoch": 1.74609375, "grad_norm": 0.038517843931913376, "learning_rate": 1.4017370040713884e-05, "loss": 0.0003440188302192837, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00034, "step": 447, "tokens/total": 13529632, "tokens/train_per_sec_per_gpu": 33.61, "tokens/trainable": 204376} +{"epoch": 1.75, "grad_norm": 0.019700370728969574, "learning_rate": 1.3898329670629645e-05, "loss": 0.00021027779439464211, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00021, "step": 448, "tokens/total": 13560080, "tokens/train_per_sec_per_gpu": 32.3, "tokens/trainable": 204821} +{"epoch": 1.75390625, "grad_norm": 0.0006826155004091561, "learning_rate": 1.3780999708818058e-05, "loss": 1.4103720786806662e-05, "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": 33.92, "tokens/trainable": 205271} +{"epoch": 1.7578125, "grad_norm": 0.008357573300600052, "learning_rate": 1.3665385037857758e-05, "loss": 9.161404886981472e-05, "memory/device_reserved (GiB)": 34.92, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00009, "step": 450, "tokens/total": 13620960, "tokens/train_per_sec_per_gpu": 31.24, "tokens/trainable": 205723} +{"epoch": 1.76171875, "grad_norm": 0.01647551730275154, "learning_rate": 1.3551490468947126e-05, "loss": 0.00017527706222608685, "memory/device_reserved (GiB)": 35.04, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00018, "step": 451, "tokens/total": 13651344, "tokens/train_per_sec_per_gpu": 32.75, "tokens/trainable": 206169} +{"epoch": 1.765625, "grad_norm": 0.03474664315581322, "learning_rate": 1.3439320741704075e-05, "loss": 0.0004909664858132601, "memory/device_reserved (GiB)": 35.04, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00049, "step": 452, "tokens/total": 13681792, "tokens/train_per_sec_per_gpu": 33.01, "tokens/trainable": 206592} +{"epoch": 1.76953125, "grad_norm": 0.005533512681722641, "learning_rate": 1.3328880523968808e-05, "loss": 8.479113603243604e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00008, "step": 453, "tokens/total": 13712256, "tokens/train_per_sec_per_gpu": 33.28, "tokens/trainable": 207055} +{"epoch": 1.7734375, "grad_norm": 0.028869982808828354, "learning_rate": 1.3220174411609587e-05, "loss": 7.264575106091797e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.00007, "step": 454, "tokens/total": 13742128, "tokens/train_per_sec_per_gpu": 34.73, "tokens/trainable": 207494} +{"epoch": 1.77734375, "grad_norm": 0.0076578096486628056, "learning_rate": 1.3113206928331471e-05, "loss": 2.4288401618832722e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 455, "tokens/total": 13772160, "tokens/train_per_sec_per_gpu": 33.22, "tokens/trainable": 207956} +{"epoch": 1.78125, "grad_norm": 0.0006251306622289121, "learning_rate": 1.300798252548806e-05, "loss": 1.4033696061233059e-05, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00001, "step": 456, "tokens/total": 13802448, "tokens/train_per_sec_per_gpu": 35.89, "tokens/trainable": 208408} +{"epoch": 1.78515625, "grad_norm": 0.0268402099609375, "learning_rate": 1.2904505581896265e-05, "loss": 0.00014731910778209567, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00015, "step": 457, "tokens/total": 13832880, "tokens/train_per_sec_per_gpu": 36.37, "tokens/trainable": 208889} +{"epoch": 1.7890625, "grad_norm": 0.060711171478033066, "learning_rate": 1.2802780403654082e-05, "loss": 0.00020443923131097108, "memory/device_reserved (GiB)": 35.64, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.0002, "step": 458, "tokens/total": 13862880, "tokens/train_per_sec_per_gpu": 34.44, "tokens/trainable": 209336} +{"epoch": 1.79296875, "grad_norm": 0.12638092041015625, "learning_rate": 1.2702811223961408e-05, "loss": 0.0009473193786107004, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00095, "step": 459, "tokens/total": 13893440, "tokens/train_per_sec_per_gpu": 33.6, "tokens/trainable": 209797} +{"epoch": 1.796875, "grad_norm": 0.0015253991587087512, "learning_rate": 1.2604602202943861e-05, "loss": 2.2379259462468326e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00002, "step": 460, "tokens/total": 13923824, "tokens/train_per_sec_per_gpu": 36.49, "tokens/trainable": 210240} +{"epoch": 1.80078125, "grad_norm": 0.013747433200478554, "learning_rate": 1.2508157427479686e-05, "loss": 9.997214510804042e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0001, "step": 461, "tokens/total": 13954208, "tokens/train_per_sec_per_gpu": 35.22, "tokens/trainable": 210702} +{"epoch": 1.8046875, "grad_norm": 0.08502306789159775, "learning_rate": 1.2413480911029655e-05, "loss": 0.0007195118232630193, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00072, "step": 462, "tokens/total": 13984784, "tokens/train_per_sec_per_gpu": 31.78, "tokens/trainable": 211140} +{"epoch": 1.80859375, "grad_norm": 0.016438405960798264, "learning_rate": 1.2320576593470082e-05, "loss": 9.172871796181425e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00009, "step": 463, "tokens/total": 14014896, "tokens/train_per_sec_per_gpu": 32.93, "tokens/trainable": 211575} +{"epoch": 1.8125, "grad_norm": 0.09976232051849365, "learning_rate": 1.2229448340928828e-05, "loss": 0.000811650650575757, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00081, "step": 464, "tokens/total": 14045312, "tokens/train_per_sec_per_gpu": 39.32, "tokens/trainable": 212083} +{"epoch": 1.81640625, "grad_norm": 0.00435422919690609, "learning_rate": 1.2140099945624458e-05, "loss": 2.5664587155915797e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00003, "step": 465, "tokens/total": 14075840, "tokens/train_per_sec_per_gpu": 36.89, "tokens/trainable": 212558} +{"epoch": 1.8203125, "grad_norm": 0.0026768327224999666, "learning_rate": 1.205253512570841e-05, "loss": 3.4629334550118074e-05, "memory/device_reserved (GiB)": 35.97, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 466, "tokens/total": 14106016, "tokens/train_per_sec_per_gpu": 33.17, "tokens/trainable": 213014} +{"epoch": 1.82421875, "grad_norm": 0.0026948358863592148, "learning_rate": 1.1966757525110255e-05, "loss": 4.344901753938757e-05, "memory/device_reserved (GiB)": 37.99, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00004, "step": 467, "tokens/total": 14136512, "tokens/train_per_sec_per_gpu": 34.3, "tokens/trainable": 213465} +{"epoch": 1.828125, "grad_norm": 0.006161310710012913, "learning_rate": 1.1882770713386095e-05, "loss": 4.4334883568808436e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00004, "step": 468, "tokens/total": 14166832, "tokens/train_per_sec_per_gpu": 36.92, "tokens/trainable": 213940} +{"epoch": 1.83203125, "grad_norm": 0.010646386072039604, "learning_rate": 1.180057818556998e-05, "loss": 2.2965485186432488e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00002, "step": 469, "tokens/total": 14197360, "tokens/train_per_sec_per_gpu": 35.38, "tokens/trainable": 214415} +{"epoch": 1.8359375, "grad_norm": 0.0007588419830426574, "learning_rate": 1.1720183362028494e-05, "loss": 1.3413812666840386e-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": 34.61, "tokens/trainable": 214880} +{"epoch": 1.83984375, "grad_norm": 0.017162201926112175, "learning_rate": 1.1641589588318387e-05, "loss": 0.0001371123653370887, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00014, "step": 471, "tokens/total": 14257888, "tokens/train_per_sec_per_gpu": 33.74, "tokens/trainable": 215334} +{"epoch": 1.84375, "grad_norm": 0.0008721952908672392, "learning_rate": 1.1564800135047418e-05, "loss": 1.1102063581347466e-05, "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": 35.27, "tokens/trainable": 215822} +{"epoch": 1.84765625, "grad_norm": 0.029533211141824722, "learning_rate": 1.148981819773816e-05, "loss": 0.0002879542880691588, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00029, "step": 473, "tokens/total": 14318448, "tokens/train_per_sec_per_gpu": 31.18, "tokens/trainable": 216239} +{"epoch": 1.8515625, "grad_norm": 0.0012399809202179313, "learning_rate": 1.1416646896695086e-05, "loss": 1.8869191990233958e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00002, "step": 474, "tokens/total": 14348912, "tokens/train_per_sec_per_gpu": 33.67, "tokens/trainable": 216691} +{"epoch": 1.85546875, "grad_norm": 0.22101251780986786, "learning_rate": 1.1345289276874717e-05, "loss": 0.0024789171293377876, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00248, "step": 475, "tokens/total": 14379440, "tokens/train_per_sec_per_gpu": 33.41, "tokens/trainable": 217162} +{"epoch": 1.859375, "grad_norm": 0.004043657332658768, "learning_rate": 1.1275748307758873e-05, "loss": 4.4153628550702706e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00004, "step": 476, "tokens/total": 14409840, "tokens/train_per_sec_per_gpu": 36.12, "tokens/trainable": 217638} +{"epoch": 1.86328125, "grad_norm": 0.022147774696350098, "learning_rate": 1.1208026883231147e-05, "loss": 0.00015884214371908456, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00016, "step": 477, "tokens/total": 14440256, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 218053} +{"epoch": 1.8671875, "grad_norm": 0.0010556846391409636, "learning_rate": 1.1142127821456433e-05, "loss": 1.8117327272193506e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 478, "tokens/total": 14470704, "tokens/train_per_sec_per_gpu": 34.42, "tokens/trainable": 218503} +{"epoch": 1.87109375, "grad_norm": 0.019303888082504272, "learning_rate": 1.1078053864763674e-05, "loss": 0.0001503409876022488, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00015, "step": 479, "tokens/total": 14501072, "tokens/train_per_sec_per_gpu": 36.47, "tokens/trainable": 218964} +{"epoch": 1.875, "grad_norm": 0.008807498030364513, "learning_rate": 1.1015807679531756e-05, "loss": 6.010168726788834e-05, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00006, "step": 480, "tokens/total": 14531200, "tokens/train_per_sec_per_gpu": 35.51, "tokens/trainable": 219412} +{"epoch": 1.87890625, "grad_norm": 0.0371716283261776, "learning_rate": 1.0955391856078528e-05, "loss": 0.00035582270356826484, "memory/device_reserved (GiB)": 38.63, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00036, "step": 481, "tokens/total": 14561152, "tokens/train_per_sec_per_gpu": 38.62, "tokens/trainable": 219886} +{"epoch": 1.8828125, "grad_norm": 0.0007208790630102158, "learning_rate": 1.0896808908553007e-05, "loss": 1.2405967027007136e-05, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00001, "step": 482, "tokens/total": 14591264, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 220343} +{"epoch": 1.88671875, "grad_norm": 0.019832292571663857, "learning_rate": 1.0840061274830763e-05, "loss": 0.0002267304516863078, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00023, "step": 483, "tokens/total": 14621568, "tokens/train_per_sec_per_gpu": 35.53, "tokens/trainable": 220825} +{"epoch": 1.890625, "grad_norm": 0.0017411233857274055, "learning_rate": 1.0785151316412473e-05, "loss": 2.01847760763485e-05, "memory/device_reserved (GiB)": 37.77, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00002, "step": 484, "tokens/total": 14652064, "tokens/train_per_sec_per_gpu": 32.83, "tokens/trainable": 221282} +{"epoch": 1.89453125, "grad_norm": 0.0007287322077900171, "learning_rate": 1.0732081318325639e-05, "loss": 9.605650120647624e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00001, "step": 485, "tokens/total": 14682608, "tokens/train_per_sec_per_gpu": 38.26, "tokens/trainable": 221780} +{"epoch": 1.8984375, "grad_norm": 0.0007819020538590848, "learning_rate": 1.0680853489029501e-05, "loss": 1.6115096514113247e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 486, "tokens/total": 14713168, "tokens/train_per_sec_per_gpu": 31.89, "tokens/trainable": 222210} +{"epoch": 1.90234375, "grad_norm": 0.014062542468309402, "learning_rate": 1.0631469960323152e-05, "loss": 0.00013912066060584038, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00014, "step": 487, "tokens/total": 14743536, "tokens/train_per_sec_per_gpu": 29.67, "tokens/trainable": 222646} +{"epoch": 1.90625, "grad_norm": 0.13683916628360748, "learning_rate": 1.0583932787256783e-05, "loss": 0.0015979098388925195, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.0016, "step": 488, "tokens/total": 14773728, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 223066} +{"epoch": 1.91015625, "grad_norm": 0.5263968706130981, "learning_rate": 1.0538243948046206e-05, "loss": 0.00542115792632103, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00544, "step": 489, "tokens/total": 14803744, "tokens/train_per_sec_per_gpu": 34.47, "tokens/trainable": 223512} +{"epoch": 1.9140625, "grad_norm": 0.03019803576171398, "learning_rate": 1.0494405343990523e-05, "loss": 0.0002152348606614396, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00022, "step": 490, "tokens/total": 14832112, "tokens/train_per_sec_per_gpu": 39.3, "tokens/trainable": 223956} +{"epoch": 1.91796875, "grad_norm": 0.047021783888339996, "learning_rate": 1.0452418799392985e-05, "loss": 0.00024185153597500175, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.44, "memory/max_allocated (GiB)": 33.44, "ppl": 1.00024, "step": 491, "tokens/total": 14860560, "tokens/train_per_sec_per_gpu": 36.71, "tokens/trainable": 224416} +{"epoch": 1.921875, "grad_norm": 0.0029584276489913464, "learning_rate": 1.0412286061485102e-05, "loss": 3.368399484315887e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00003, "step": 492, "tokens/total": 14890848, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 224883} +{"epoch": 1.92578125, "grad_norm": 0.005298682954162359, "learning_rate": 1.03740088003539e-05, "loss": 3.912312240572646e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00004, "step": 493, "tokens/total": 14921008, "tokens/train_per_sec_per_gpu": 35.8, "tokens/trainable": 225324} +{"epoch": 1.9296875, "grad_norm": 0.007532566320151091, "learning_rate": 1.0337588608872463e-05, "loss": 7.047319377306849e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00007, "step": 494, "tokens/total": 14951360, "tokens/train_per_sec_per_gpu": 35.23, "tokens/trainable": 225783} +{"epoch": 1.93359375, "grad_norm": 0.006395815405994654, "learning_rate": 1.0303027002633622e-05, "loss": 4.224019357934594e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00004, "step": 495, "tokens/total": 14981552, "tokens/train_per_sec_per_gpu": 36.89, "tokens/trainable": 226281} +{"epoch": 1.9375, "grad_norm": 0.0026577117387205362, "learning_rate": 1.0270325419886884e-05, "loss": 1.877904105640482e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00002, "step": 496, "tokens/total": 15011584, "tokens/train_per_sec_per_gpu": 37.11, "tokens/trainable": 226735} +{"epoch": 1.94140625, "grad_norm": 0.003066121833398938, "learning_rate": 1.0239485221478599e-05, "loss": 3.809008558164351e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00004, "step": 497, "tokens/total": 15042016, "tokens/train_per_sec_per_gpu": 36.66, "tokens/trainable": 227198} +{"epoch": 1.9453125, "grad_norm": 0.0003162898647133261, "learning_rate": 1.0210507690795292e-05, "loss": 9.668269740359392e-06, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.68, "memory/max_allocated (GiB)": 33.68, "ppl": 1.00001, "step": 498, "tokens/total": 15071984, "tokens/train_per_sec_per_gpu": 31.4, "tokens/trainable": 227623} +{"epoch": 1.94921875, "grad_norm": 0.035544853657484055, "learning_rate": 1.0183394033710305e-05, "loss": 0.00024893757654353976, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00025, "step": 499, "tokens/total": 15102192, "tokens/train_per_sec_per_gpu": 32.48, "tokens/trainable": 228074} +{"epoch": 1.953125, "grad_norm": 0.005648774094879627, "learning_rate": 1.0158145378533583e-05, "loss": 4.911683936370537e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00005, "step": 500, "tokens/total": 15132560, "tokens/train_per_sec_per_gpu": 35.87, "tokens/trainable": 228541} +{"epoch": 1.95703125, "grad_norm": 0.0024769131559878588, "learning_rate": 1.0134762775964726e-05, "loss": 2.570717151684221e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00003, "step": 501, "tokens/total": 15163152, "tokens/train_per_sec_per_gpu": 38.07, "tokens/trainable": 229031} +{"epoch": 1.9609375, "grad_norm": 0.005817459896206856, "learning_rate": 1.0113247199049278e-05, "loss": 3.874715912388638e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00004, "step": 502, "tokens/total": 15193456, "tokens/train_per_sec_per_gpu": 39.01, "tokens/trainable": 229512} +{"epoch": 1.96484375, "grad_norm": 0.213705375790596, "learning_rate": 1.0093599543138205e-05, "loss": 0.0010556308552622795, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00106, "step": 503, "tokens/total": 15223680, "tokens/train_per_sec_per_gpu": 34.53, "tokens/trainable": 229989} +{"epoch": 1.96875, "grad_norm": 0.0057237111032009125, "learning_rate": 1.0075820625850675e-05, "loss": 2.9780632758047432e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00003, "step": 504, "tokens/total": 15254336, "tokens/train_per_sec_per_gpu": 36.15, "tokens/trainable": 230447} +{"epoch": 1.97265625, "grad_norm": 0.017064962536096573, "learning_rate": 1.0059911187040013e-05, "loss": 0.0001428169198334217, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00014, "step": 505, "tokens/total": 15284496, "tokens/train_per_sec_per_gpu": 34.6, "tokens/trainable": 230879} +{"epoch": 1.9765625, "grad_norm": 0.004507618024945259, "learning_rate": 1.0045871888762893e-05, "loss": 1.688709744485095e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00002, "step": 506, "tokens/total": 15314976, "tokens/train_per_sec_per_gpu": 34.46, "tokens/trainable": 231348} +{"epoch": 1.98046875, "grad_norm": 0.0015678065828979015, "learning_rate": 1.003370331525184e-05, "loss": 2.9251012165332213e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00003, "step": 507, "tokens/total": 15345280, "tokens/train_per_sec_per_gpu": 35.63, "tokens/trainable": 231819} +{"epoch": 1.984375, "grad_norm": 0.0032501902896910906, "learning_rate": 1.002340597289085e-05, "loss": 4.6847708290442824e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00005, "step": 508, "tokens/total": 15375424, "tokens/train_per_sec_per_gpu": 34.94, "tokens/trainable": 232269} +{"epoch": 1.98828125, "grad_norm": 0.003683724207803607, "learning_rate": 1.0014980290194387e-05, "loss": 3.591051427065395e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00004, "step": 509, "tokens/total": 15405872, "tokens/train_per_sec_per_gpu": 34.4, "tokens/trainable": 232716} +{"epoch": 1.9921875, "grad_norm": 0.0030913210939615965, "learning_rate": 1.0008426617789489e-05, "loss": 2.7738915377995e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 510, "tokens/total": 15436128, "tokens/train_per_sec_per_gpu": 36.41, "tokens/trainable": 233178} +{"epoch": 1.99609375, "grad_norm": 0.04374156519770622, "learning_rate": 1.0003745228401215e-05, "loss": 4.1764309571590275e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00004, "step": 511, "tokens/total": 15466464, "tokens/train_per_sec_per_gpu": 36.21, "tokens/trainable": 233647} +{"epoch": 2.0, "grad_norm": 0.00487576425075531, "learning_rate": 1.0000936316841296e-05, "loss": 3.3267013350268826e-05, "memory/device_reserved (GiB)": 38.44, "memory/max_active (GiB)": 33.67, "memory/max_allocated (GiB)": 33.67, "ppl": 1.00003, "step": 512, "tokens/total": 15496368, "tokens/train_per_sec_per_gpu": 29.68, "tokens/trainable": 234060}