diff --git a/checkpoint-3500/latest b/checkpoint-3500/latest index 2a79fdc19587e6bc9de060e90633f3a151b04516..f3af93198d9ca215e3991f85bf3781ae5dfe71f3 100644 --- a/checkpoint-3500/latest +++ b/checkpoint-3500/latest @@ -1 +1 @@ -global_step2000 \ No newline at end of file +global_step3500 \ No newline at end of file diff --git a/checkpoint-3500/model-00001-of-00004.safetensors b/checkpoint-3500/model-00001-of-00004.safetensors index 320da25ceeb9ec7a9775c01e17cb6fadf2aff184..f91755ed2fdc9ff9bc885b1d2e65b5e6860f14ca 100644 --- a/checkpoint-3500/model-00001-of-00004.safetensors +++ b/checkpoint-3500/model-00001-of-00004.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c1426d4f131e32848765a90f9fa8338f4a644ce103c6291b705482638c51220f +oid sha256:a6a7e13a55cda518bc5eff6f15b6994a133dd1bbfbf1ecaf5d2b2f8f5dd57eb8 size 4994639360 diff --git a/checkpoint-3500/model-00002-of-00004.safetensors b/checkpoint-3500/model-00002-of-00004.safetensors index 9af6b4dc97d9972c1d82956698d70ec5f8b54ada..a2bc4e7a1f18b9a8e794d9e145f89573600fc841 100644 --- a/checkpoint-3500/model-00002-of-00004.safetensors +++ b/checkpoint-3500/model-00002-of-00004.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7a4a3a870a4f515f1e895fba8f97fa5f7dfedc3ba36a5e9b8375969b6cc419e6 +oid sha256:8e7baf92a86aaad05419878dfdb73292f95dfe93d180143cdbd9493748289394 size 4999802600 diff --git a/checkpoint-3500/model-00003-of-00004.safetensors b/checkpoint-3500/model-00003-of-00004.safetensors index 9fc8fd90df40ed3eff4258909f9c188b8a5dd288..7acecb31de7d207dd9c29387cf007258af980bc8 100644 --- a/checkpoint-3500/model-00003-of-00004.safetensors +++ b/checkpoint-3500/model-00003-of-00004.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:622f7ca55870a26d128c5091168c64b11e49be9c78bc330468525f160881af0d +oid sha256:20c2ce05978d4aee32acea8e5a3eb9b00882d4b772d0e09666e3c0c87cbb34c5 size 4999827272 diff --git a/checkpoint-3500/model-00004-of-00004.safetensors b/checkpoint-3500/model-00004-of-00004.safetensors index d1260e1bc0252f95e460218dba8e04e01ce5f671..4683f0aea5c5c45a5f662172c41259f0f96b9e8f 100644 --- a/checkpoint-3500/model-00004-of-00004.safetensors +++ b/checkpoint-3500/model-00004-of-00004.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:4486a62b3444b6671b2061cf92ac1ab7980b20f93d60ee24a580514d5583e900 +oid sha256:873bf38ff5050cd2515c900e56ada7ba954fbee3c01d1384f2a9d540133eac2c size 1937667242 diff --git a/checkpoint-3500/rng_state_0.pth b/checkpoint-3500/rng_state_0.pth index 5cdc5b0e0b80e2abd88f1761be343e9f6a122f5e..711b2ad5bd7792fe6ea586b3804e75af9c9fd78b 100644 --- a/checkpoint-3500/rng_state_0.pth +++ b/checkpoint-3500/rng_state_0.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:8bb40a4458c9e72a7f1d8c85b3ae935d07fad4e159a23ce47148de5aa7305ac9 +oid sha256:98c9a68d367528c32d9e5c28a0059484146710b0459b1cc5a78f503d862d2bb6 size 16389 diff --git a/checkpoint-3500/rng_state_1.pth b/checkpoint-3500/rng_state_1.pth index 5d8c1a86812cd74d0a3a51eea2458660f4626728..d7bbd1546203e239d2f127e464eb6ab5fe0a266b 100644 --- a/checkpoint-3500/rng_state_1.pth +++ b/checkpoint-3500/rng_state_1.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7a861b5eed890056d53efc68d7d171c187f7fb95c97fba3dd075a0a715d87399 +oid sha256:574d4429f18b7bbcd409bf5fa01db174e1a98f3c663e74efd8233bb63ae37e55 size 16389 diff --git a/checkpoint-3500/rng_state_10.pth b/checkpoint-3500/rng_state_10.pth index 3a2a53c60b5114a610062091b698e070b5702a47..cbe48e306130a30dd480c727fe9cdac3b30f7f3c 100644 --- a/checkpoint-3500/rng_state_10.pth +++ b/checkpoint-3500/rng_state_10.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7268593a09f3f04c55240873078ae152b3f5c892cacab1ba48e57ec65a291281 +oid sha256:5621b8c68425c7c03108f6b5b66c13298a96042629813f7e1865b1f7a00aaf58 size 16404 diff --git a/checkpoint-3500/rng_state_11.pth b/checkpoint-3500/rng_state_11.pth index 3cf90bae730002d121c880cfef5faa23aecc5ea4..230e47c11cc418ea24f26a6455587edebfc544f6 100644 --- a/checkpoint-3500/rng_state_11.pth +++ b/checkpoint-3500/rng_state_11.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:29352fe9566b063ec4ebef7fc654e2ded4262d3a4f9db7b0223feee24869096d +oid sha256:a85a83468a2de432e39bda54cc32ff345770d18f5d69633811f7fbd71c720deb size 16404 diff --git a/checkpoint-3500/rng_state_12.pth b/checkpoint-3500/rng_state_12.pth index d96223384ba05de0d56f8def700cf304f079fc30..86a8d2079de70916482fd846040c6c9b03eb03f6 100644 --- a/checkpoint-3500/rng_state_12.pth +++ b/checkpoint-3500/rng_state_12.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2d09ac763a1adcef2bf4df49f00f7f7871a152f37ee1628b3b0d6307dc155df8 +oid sha256:c89c12e4df8b4cf54de85737ba6d9ceb27b90082115a7c58a2fdd586ed1549de size 16404 diff --git a/checkpoint-3500/rng_state_13.pth b/checkpoint-3500/rng_state_13.pth index 12b203c21f30a35bc95d6e25623c5bf40767a52f..d98b066910670c910b140548ab3364d225594c75 100644 --- a/checkpoint-3500/rng_state_13.pth +++ b/checkpoint-3500/rng_state_13.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:632f157c6356b0cc0a96d17184cc70bbee6659ec11b5a40aaff4366dae485134 +oid sha256:74ed49a68d6c99371f24815c9f2858fc7b5e2356f9367eaa0da6b469b3f8f163 size 16404 diff --git a/checkpoint-3500/rng_state_14.pth b/checkpoint-3500/rng_state_14.pth index 0eaccb4a6eade742fb7018a54743ac7718ce46ef..ebf142c4f6322d25e3e02a6c262af996fc74bf22 100644 --- a/checkpoint-3500/rng_state_14.pth +++ b/checkpoint-3500/rng_state_14.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:09ca89da7bffc60ad5da329197beb7a01870d959bdbbc4e60074e0d722140dca +oid sha256:a2a201b69199a4b625ddfb27d3cb541d55182d1549771bb3341bb0a57a46a407 size 16404 diff --git a/checkpoint-3500/rng_state_15.pth b/checkpoint-3500/rng_state_15.pth index e78f14f2c675b145e569742943c07906caf6ecfe..80be9b286bf4be62f6be3c4c209dd5c6f616bc3f 100644 --- a/checkpoint-3500/rng_state_15.pth +++ b/checkpoint-3500/rng_state_15.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:d598722cdad23db99108b57a480e2cc5a7490492e3d8adc148c1ffb1218107b5 +oid sha256:accc08ab8b3a671a16c273d869807f6a83122834b981be6c23f4a64110d64b7c size 16404 diff --git a/checkpoint-3500/rng_state_16.pth b/checkpoint-3500/rng_state_16.pth index da0a45603c8f934268f0eca160fa7e239b414c1b..f972caae35d4e59bbf381b2f01ef044b685c4c7b 100644 --- a/checkpoint-3500/rng_state_16.pth +++ b/checkpoint-3500/rng_state_16.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1c0917ef1d32a951ac46f9d7aef4ce8775bc382018bb8c5e141ef696055c0661 +oid sha256:91c3d3ca6309abbc621bc5ce5d7a744a4ab98732a29575cd9300a747bfc643f7 size 16404 diff --git a/checkpoint-3500/rng_state_17.pth b/checkpoint-3500/rng_state_17.pth index cf92642dcfcc7d12bb4a1b6e0a4ae43ea76a37a8..05e9a07a89c7bc81ab647330e7b6e7939c9f0956 100644 --- a/checkpoint-3500/rng_state_17.pth +++ b/checkpoint-3500/rng_state_17.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:d684ae7b60674391a4725bc249c0b429fce908d09b39b5a304dc1081778831f3 +oid sha256:ecfc8dfa1d635b57a52f87f3f6a922bcf08d1d03cbd7a6483b79fe0cf3b18e74 size 16404 diff --git a/checkpoint-3500/rng_state_18.pth b/checkpoint-3500/rng_state_18.pth index 3465b975e88e3e1257bbc5f27956cccc67463e9a..7e3851fbbda0314824e04a859e95cfd68d4e5d47 100644 --- a/checkpoint-3500/rng_state_18.pth +++ b/checkpoint-3500/rng_state_18.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:34ce6c2c3d76c1b842de274d283046b5e55a1c9fac25166137013e62ed16ef7e +oid sha256:cfcf04d7692ba3ab58e525a2ae70c310b41a273556c2b628f6c0f8b3089db30c size 16404 diff --git a/checkpoint-3500/rng_state_19.pth b/checkpoint-3500/rng_state_19.pth index e860544bb2f216245b8e4e0ce4b095ae907725bc..0ba41d770acf9f2a77cb720e034b00a69506b6dc 100644 --- a/checkpoint-3500/rng_state_19.pth +++ b/checkpoint-3500/rng_state_19.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2c4ffaed3534aee2a29d3d752d8247290d87b829aa0adf367cb266aff1ae471d +oid sha256:497c30c719f03409c7e9f887e0d50d4e432546da644ddc19d4750f15f96e8134 size 16404 diff --git a/checkpoint-3500/rng_state_2.pth b/checkpoint-3500/rng_state_2.pth index 1ea629048a685184cc3c37c7f680a3b29393f630..489d7c124823698a88dcf686e94f37f68ed8f428 100644 --- a/checkpoint-3500/rng_state_2.pth +++ b/checkpoint-3500/rng_state_2.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:be9d0b12d228a0a1524f873c06dba477d4c0ca109104ce01b91b6e13e70bff3c +oid sha256:02401b9686e4e9124a1894f7f5744443d8b8961dabdac7af763c437f2f3fc9f1 size 16389 diff --git a/checkpoint-3500/rng_state_20.pth b/checkpoint-3500/rng_state_20.pth index 9f300fcd4193e476043a2b57479dcad307d0cccd..0415ac3dd6bc0956757b8cd5d6065acb3ad3d717 100644 --- a/checkpoint-3500/rng_state_20.pth +++ b/checkpoint-3500/rng_state_20.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:8fd064ef0fabfe96bf7db3aab96c9e6cc05e0a80e2c06deaf7677537dc8ab3b5 +oid sha256:b4721c7d565df421745b6df4502680f38e60451235e3bff2bcba9af0c4e046fc size 16404 diff --git a/checkpoint-3500/rng_state_21.pth b/checkpoint-3500/rng_state_21.pth index cba9f6fd1eba4fb708ec275ba13ce6d3e91be449..fba752aeca5d6804124074e0f31fd455f1b046ec 100644 --- a/checkpoint-3500/rng_state_21.pth +++ b/checkpoint-3500/rng_state_21.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:47b41fcfee855dfd7d255da222b343c6b25e3d8440e5ab787860964c390dea9f +oid sha256:3cd096e6ce97d382528cf85e3c2d7f4fb7b471bb6cb714bef730e0521da73e3b size 16404 diff --git a/checkpoint-3500/rng_state_22.pth b/checkpoint-3500/rng_state_22.pth index 7d70a5e933b8e6bd842ed91fbaef96fcaa1a7e76..2cf35d223cd9dd8fd3f2a086c19b79c52b3c6281 100644 --- a/checkpoint-3500/rng_state_22.pth +++ b/checkpoint-3500/rng_state_22.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b2493113f0f3bec5ea239e877f553fde077c5710173f1e905ca1f23f7836e531 +oid sha256:b4670f0021f5515489da414563cfc63f457277319b2dfb071e350f563dedddf5 size 16404 diff --git a/checkpoint-3500/rng_state_23.pth b/checkpoint-3500/rng_state_23.pth index bf2d04aaf0dc489ae95f9b201bc7802899d6ddca..9da1b37a6755e2815267084c1edfc5419b8ecf6a 100644 --- a/checkpoint-3500/rng_state_23.pth +++ b/checkpoint-3500/rng_state_23.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:0d39456dc7716cc60babbb70a8b7f6605e1f72946c58c2c836481c29417cfb08 +oid sha256:42ead2db41e126606e8c45ebde995efc6004e5a74989075df740ed1b091df876 size 16404 diff --git a/checkpoint-3500/rng_state_24.pth b/checkpoint-3500/rng_state_24.pth index 08027cfd5f5509625221f170e9bfc76edd0e9786..52851ab9db31de0f5aa5beeb984f54a9d034b157 100644 --- a/checkpoint-3500/rng_state_24.pth +++ b/checkpoint-3500/rng_state_24.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:a5e2152a2792f48ffbebc9dc15af2ed819dfc56f04f0611e052163100b25ab62 +oid sha256:2bb91d829cb4d67287c01c7152c6012180ad9e4cc8c38b8455c6e7a4e95d47bb size 16404 diff --git a/checkpoint-3500/rng_state_25.pth b/checkpoint-3500/rng_state_25.pth index ab5c69ca4f3724f3a8ba93cedc9f9b38d30328b5..c3cddf96c2ab92ee0c1441eb2fb59d4983c47a23 100644 --- a/checkpoint-3500/rng_state_25.pth +++ b/checkpoint-3500/rng_state_25.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:361af5af89e7a00d67fe830cc0d764fe88b002d8f9f3bcf11ff3abbd73eb35e1 +oid sha256:40b728aa8401ac27d0688a582b6d7030f4d61629557a7a0759bfa51e2b5d5ec9 size 16404 diff --git a/checkpoint-3500/rng_state_26.pth b/checkpoint-3500/rng_state_26.pth index 948f9f8c30d58aeb603c15d938a0a10b6bd3453a..96c345423c3f9d4434e6570ccf01185f28374fa0 100644 --- a/checkpoint-3500/rng_state_26.pth +++ b/checkpoint-3500/rng_state_26.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2896b5f3f08fdcee9cd91ee98455c01d6620e978c83e20a1dcfccbd73e37e2c1 +oid sha256:a0b932c344010ebb8c6bab7784a9899dfd466ab5376108f642852c54386b20f8 size 16404 diff --git a/checkpoint-3500/rng_state_27.pth b/checkpoint-3500/rng_state_27.pth index 824ab47c2e7b747225b4bae767f3a2509bc34802..d7273ab80abacbcf9daabbbc35e8e659a783d4fe 100644 --- a/checkpoint-3500/rng_state_27.pth +++ b/checkpoint-3500/rng_state_27.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:4b1510b51356e22b62a1cfe58911d5de7cfdf8348dfda14be59658f0178290dc +oid sha256:696fc950a4b8a557a7a4a93dc07ff0cb2d77ba990a41d9c6b406d32244d06ce4 size 16404 diff --git a/checkpoint-3500/rng_state_28.pth b/checkpoint-3500/rng_state_28.pth index aacad8a0455d1cd2fdea73bf9acb8c41b66a4a1c..c4e5406a3c4c0de9fee63df22be6542695d9faa8 100644 --- a/checkpoint-3500/rng_state_28.pth +++ b/checkpoint-3500/rng_state_28.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1d304c97fb350b997a8dbdbb065f0e3e3f7fa6809a02ad7da2808ca98b2c2211 +oid sha256:05602054e9dba9b0313752945e9b1d8874031c61c720968798095b1c168c201b size 16404 diff --git a/checkpoint-3500/rng_state_29.pth b/checkpoint-3500/rng_state_29.pth index 20f866dc4ee25cd35f70f34fda5f55f3685ad4f3..33b9b945cb26318b0272a83ac990a96ce4530773 100644 --- a/checkpoint-3500/rng_state_29.pth +++ b/checkpoint-3500/rng_state_29.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:be0ac68aaea236f1339f3495a9c29c030ce505fda0e6b5943010cd245d707c3d +oid sha256:2c07e2a06d965cb2ed1b49e84d5faedab0390a8030950747873bdf9394b8714f size 16404 diff --git a/checkpoint-3500/rng_state_3.pth b/checkpoint-3500/rng_state_3.pth index f4052eac64454fa156e5ed5a97c247b222973295..f8b6d364c88e63bbe989bff4d56ab2aa80c85309 100644 --- a/checkpoint-3500/rng_state_3.pth +++ b/checkpoint-3500/rng_state_3.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:17147d81339261eba40ea98a307d42d534c9791e014b827c6916ae73ee0ac517 +oid sha256:00127033676eacf74c724bcbeb6972c3d9dec0870417ed85f9a2dde067033b44 size 16389 diff --git a/checkpoint-3500/rng_state_30.pth b/checkpoint-3500/rng_state_30.pth index c9ab85a24df7b83a1c56e2e145d8b5ecce85eadf..aba574fd6bfacc28d985da0d9daa59591a1ca691 100644 --- a/checkpoint-3500/rng_state_30.pth +++ b/checkpoint-3500/rng_state_30.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:85b806e8d04ec3b68848b5ba7b0c284395b073708125df909f35903c9621b70b +oid sha256:f64399a212323e9bee7c6d9ba9f56b6952548173f21fb67362e0466c9c2ad8fc size 16404 diff --git a/checkpoint-3500/rng_state_31.pth b/checkpoint-3500/rng_state_31.pth index 4f9f82004f494e1a74fc31d2458b1a07644ca43f..1a1ec1165955500dbd339fe30c45617a84a38adf 100644 --- a/checkpoint-3500/rng_state_31.pth +++ b/checkpoint-3500/rng_state_31.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:401ae69e52a854d261a6f4d2bf0770994cd12a3900b20eaceac0b045867866bf +oid sha256:5d26753195ced20755679182dc934184a51db9ad812f9330be44bd4e4640c380 size 16404 diff --git a/checkpoint-3500/rng_state_32.pth b/checkpoint-3500/rng_state_32.pth index 3055c2382e670fb31dbc457c35702dddb1d089a2..5b4e060ae2310c4cd7a6e5b08e3ca16b2380dd44 100644 --- a/checkpoint-3500/rng_state_32.pth +++ b/checkpoint-3500/rng_state_32.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:5a3cf2936b9b1f184e96be398737b2c0cec89ae6159cf5247956369a3a3684f7 +oid sha256:3140681bb6b0f49765bfea2908a727122ee986c39f55a424f6712b89630461ac size 16404 diff --git a/checkpoint-3500/rng_state_33.pth b/checkpoint-3500/rng_state_33.pth index 78c268c20f3b499483686437e48a7156dda812d2..c64e86b485fb1add4fab95a4a8969ded3b2a54c5 100644 --- a/checkpoint-3500/rng_state_33.pth +++ b/checkpoint-3500/rng_state_33.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e5f6cfeae0e4bc20f7c31c244ea7569748d148e3e8fd39f1b3ae745b8a545500 +oid sha256:dc60a5d7ddacf0cdcebb4bdc85eeb5b35fef00174d2f0af2d02e0e6de582613c size 16404 diff --git a/checkpoint-3500/rng_state_34.pth b/checkpoint-3500/rng_state_34.pth index b47872a86bb5ab4395c82fcaad3ca2af1f4b1a13..89693ce07caf3a11737c7fa62de72bcaefa204d7 100644 --- a/checkpoint-3500/rng_state_34.pth +++ b/checkpoint-3500/rng_state_34.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ce338f450e8fc88d15b0ad4a5bbbca3f8c9ae8fe50143be9935d118e059b8f22 +oid sha256:b3d47ca2458151418bc12b5f75c3ddc3430d6818605c92d09342743b96b0df19 size 16404 diff --git a/checkpoint-3500/rng_state_35.pth b/checkpoint-3500/rng_state_35.pth index 03080c26bdfeb537b25eb418b6d0e8acde160c1e..c45566729fdbdf7047393470a9dff4e100ecc338 100644 --- a/checkpoint-3500/rng_state_35.pth +++ b/checkpoint-3500/rng_state_35.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:78d25041bca3399da01a36a5c8e50a264b5d1320f16b8074f6bd79218ef12cbe +oid sha256:68f1b332efe85e73418f950df15485e62c53d4f6b68a1eeaee497ca5cdb3326c size 16404 diff --git a/checkpoint-3500/rng_state_36.pth b/checkpoint-3500/rng_state_36.pth index 8bfd7a23190dcb5826e55a91be46ff6266f689ff..f0477f6b629849aeeecb776265977e09c9385445 100644 --- a/checkpoint-3500/rng_state_36.pth +++ b/checkpoint-3500/rng_state_36.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:059cf2df08ebaa0d3361da3547027aad2a94d888fadd052931c442ee4dca2d7b +oid sha256:4bebc47a1116a2ff24e73b4add64d88c576261623796756f0c5d9f551b37d817 size 16404 diff --git a/checkpoint-3500/rng_state_37.pth b/checkpoint-3500/rng_state_37.pth index 1f60a70cac69dbd129c98b798ee26304d4e64280..4426a142e165fb4bca8841f743c77e8c69a8414f 100644 --- a/checkpoint-3500/rng_state_37.pth +++ b/checkpoint-3500/rng_state_37.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:a0a9e0d853aec2f2ecceb9137e6bcf081c6a00b196c36308a318dc42a1cccf68 +oid sha256:acc46b27b31aa6f42aeb234ebe78cdcc143e9b551cf393c0c41a75a1a0aaf88f size 16404 diff --git a/checkpoint-3500/rng_state_38.pth b/checkpoint-3500/rng_state_38.pth index de5b36039e6a735728e545d2b5dbb614f9ad0852..c09f03b7af43fb46949283608e15582a60273b5b 100644 --- a/checkpoint-3500/rng_state_38.pth +++ b/checkpoint-3500/rng_state_38.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b5c89dd2cb380a0c3b82fc3a31bb929219a9bec664c209f7d79383a6b09421b1 +oid sha256:0130ddf0be0fe902a1ecfc0ee491ce4ba82ce9b974109856b5e5c6e1ab20a39c size 16404 diff --git a/checkpoint-3500/rng_state_39.pth b/checkpoint-3500/rng_state_39.pth index bd394282c6cf1dd123103ab2e970847ea779209a..0cdbf6aab6916ef9c888460e48cb9f854bc6a6ca 100644 --- a/checkpoint-3500/rng_state_39.pth +++ b/checkpoint-3500/rng_state_39.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:6373564133ed622e08e2e75759a6f3399285d58fc5973f24c0805960115743c9 +oid sha256:834a1cb88e334f2a2e1d7d26bc05e5b0230bff06eb41d33b00c60c999314ce8b size 16404 diff --git a/checkpoint-3500/rng_state_4.pth b/checkpoint-3500/rng_state_4.pth index 3287b042fb7d4bd0905f643705bf0a9ecc57149a..bec25ba5b4541e796d86952156d1d3fc480dcae9 100644 --- a/checkpoint-3500/rng_state_4.pth +++ b/checkpoint-3500/rng_state_4.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e1a5bf7e122f6b721cb63546e00b49e7650cf627f5723eb606e85ce130785b68 +oid sha256:08c7b0c2f7c3b09f9064fc10dea610e36f364b6b9dde954b78d1a4147a227445 size 16389 diff --git a/checkpoint-3500/rng_state_40.pth b/checkpoint-3500/rng_state_40.pth index 45e281d9f421c513f8dc9b4b4f3d8ac897694107..e3f5961fdb5f8c57e7e5ccae8c98b99e449eb57a 100644 --- a/checkpoint-3500/rng_state_40.pth +++ b/checkpoint-3500/rng_state_40.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:3578c3e778a470ceff3cbef51302636b38eb8e15b25405d89d7f5e56f0eaa139 +oid sha256:be2a9e5e7e2d56dce7d81a14871c9b9e13fe046d9f21145f94456e6cf7930ae1 size 16404 diff --git a/checkpoint-3500/rng_state_41.pth b/checkpoint-3500/rng_state_41.pth index 5e3e7031c9cb3b65d5202e0f4751dc0596ded8c7..1ea2b07f120a839a78071972554b74d68b765e34 100644 --- a/checkpoint-3500/rng_state_41.pth +++ b/checkpoint-3500/rng_state_41.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:3226819809cd8b6286935b961f74e224808d40796d04839233cee1041d22a794 +oid sha256:0ac0db4e02b266967b562db7e17cf1eeff424a701bb857618a8f64af99aa5ddd size 16404 diff --git a/checkpoint-3500/rng_state_42.pth b/checkpoint-3500/rng_state_42.pth index 1958bd09c00603a87e4f222bcf09a485a82112c2..5011976c898a56fe0186e42a3c179565d5172b29 100644 --- a/checkpoint-3500/rng_state_42.pth +++ b/checkpoint-3500/rng_state_42.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:03bc4ec6674018d9db1155ef1251d9d8e3b27cb237eaefe3cdd993a85e452d9e +oid sha256:6f92902108b766d3db1ffc14c92db94305cad20d6b9077b84dedaedcfd56c573 size 16404 diff --git a/checkpoint-3500/rng_state_43.pth b/checkpoint-3500/rng_state_43.pth index 297e668482a635b1aac3daf8725c9c48cce2a8fa..b5745715db496327e1619153a2dd80db77356200 100644 --- a/checkpoint-3500/rng_state_43.pth +++ b/checkpoint-3500/rng_state_43.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f37e3822fda8d78099a34325e419ed7fcb9d40646fce0297a7a20da92250af62 +oid sha256:a8590f120ac6963d157d452fbe1e3bc97755e212db39e27830f10a7e3e6f5c5f size 16404 diff --git a/checkpoint-3500/rng_state_44.pth b/checkpoint-3500/rng_state_44.pth index b8c8a5d8d6300f80dddeb5e7b4ae791d9b359b39..2a8c435565ef0b13c06e6f1b6501f549d36d9619 100644 --- a/checkpoint-3500/rng_state_44.pth +++ b/checkpoint-3500/rng_state_44.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:a59a1f25968c9c532e0fc91fecc681ba83524f69c9d34d3696f3d222909d4b9a +oid sha256:a86cf543f70d27d63ac31eb9572d9be5624bdf89d143fae746eee4b595e2c662 size 16404 diff --git a/checkpoint-3500/rng_state_45.pth b/checkpoint-3500/rng_state_45.pth index bef6abe620378630593206f54f99eb6a680b2f19..beb5e2beb838a9bd17a5d755d2994635acf79fe7 100644 --- a/checkpoint-3500/rng_state_45.pth +++ b/checkpoint-3500/rng_state_45.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f33cead796823f6febe4fee017a5b19819b63c77e17cc711c3268a7fa036f000 +oid sha256:f227c99f102c45fe9921f6909cd0d145c989b4689401cc9c3ef13fa2384a6037 size 16404 diff --git a/checkpoint-3500/rng_state_46.pth b/checkpoint-3500/rng_state_46.pth index e4ff855fd9a82c0d3a3a9a8f53f6b41477c74c6a..68b4b73216a01f69d337acaa20489841e1ac04ff 100644 --- a/checkpoint-3500/rng_state_46.pth +++ b/checkpoint-3500/rng_state_46.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c9364d2a25bb8745c9f073dc5d13cec02178fe83ff3012154bd7113454034838 +oid sha256:ef9c209889a91c8f2534daa81eb63027053adb49e34c326b36ced306378f63aa size 16404 diff --git a/checkpoint-3500/rng_state_47.pth b/checkpoint-3500/rng_state_47.pth index f1cf9e682e9bebd75332d580812370ef7cc597de..f67d32d4fb2a79a53ca1ae726a11668b9dcbdb39 100644 --- a/checkpoint-3500/rng_state_47.pth +++ b/checkpoint-3500/rng_state_47.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:d0f0501d05dee7631c9a279ba1041f31e6b8167a89a27b8ec35b5e6ff0817994 +oid sha256:7b87e9e80a8c65edb36bfa26582e75bd0788f7eab0f0e21315b9ecfa6b1c690a size 16404 diff --git a/checkpoint-3500/rng_state_48.pth b/checkpoint-3500/rng_state_48.pth index fe6bd07f158be5874715bb1fefe4808f3136b3a0..addbdd1fe91408336bb087b43bd17a6d8f9e3939 100644 --- a/checkpoint-3500/rng_state_48.pth +++ b/checkpoint-3500/rng_state_48.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1d246f32af02a4903af9c90dd01c8d943b4f2f5f5d547392cf0db7c86abbebad +oid sha256:9d014e96480b060e5d1802b231cdf218c0be5fd660ca704490e1529fbfbf460f size 16404 diff --git a/checkpoint-3500/rng_state_49.pth b/checkpoint-3500/rng_state_49.pth index bff107874894c9a680462a66c426029a5f1fb9e7..4545cdf1579502a636cf440c14b766b9c08f2555 100644 --- a/checkpoint-3500/rng_state_49.pth +++ b/checkpoint-3500/rng_state_49.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:022c39d08dbbccacbec1d531cd8c399bd0b60d51767ea1005697304724544f71 +oid sha256:71289a728bfa64c78f4a6066e7620305114278959bd356a9acf171d8e9e36c5c size 16404 diff --git a/checkpoint-3500/rng_state_5.pth b/checkpoint-3500/rng_state_5.pth index 0ae2ded270e43c5487acf60dbbdd099e7dce504b..46937fdf126120f5566d0f0faae17623022193c6 100644 --- a/checkpoint-3500/rng_state_5.pth +++ b/checkpoint-3500/rng_state_5.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:55967c977ef7e8692e75055c69889b6f863dc5ba4b86b572e02db26c413d17bc +oid sha256:2b4ef4fc06d101fbc79f7360cf5b6b622718d9c11c8f03db6d08bb6b5b1497fa size 16389 diff --git a/checkpoint-3500/rng_state_50.pth b/checkpoint-3500/rng_state_50.pth index cfa5f3fb5f2e034dd0023a035fab745a0c79d4e3..9917afefec37d8a12cf58d1bf89b4227d651b7b3 100644 --- a/checkpoint-3500/rng_state_50.pth +++ b/checkpoint-3500/rng_state_50.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:52d33c52f906a1b8f026ab6659f7c7835ab5db90f792f61ca987a42ae60e023d +oid sha256:63dfb3bbf82eeef63ad8d5e7142b556965cf06548e5a455269a9c587821561cb size 16404 diff --git a/checkpoint-3500/rng_state_51.pth b/checkpoint-3500/rng_state_51.pth index 6fcd72f506583caefa036f35a4aedde55e6f406f..0429e4cde014ad6595805ed9bd200c7e58b00a2c 100644 --- a/checkpoint-3500/rng_state_51.pth +++ b/checkpoint-3500/rng_state_51.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:fbe93f2e334996ba960ee92d1a58c42ba02e8c030a650a3be9c975bb5177de30 +oid sha256:34965ca3746db7ede8269665d276857a9709f09ce720e53f99eb4662381354a0 size 16404 diff --git a/checkpoint-3500/rng_state_52.pth b/checkpoint-3500/rng_state_52.pth index b3c531aa12ff4fbcae342c11380adf79f33693d1..5dcb5d1c240e39de9ce661a86c8c953d84743746 100644 --- a/checkpoint-3500/rng_state_52.pth +++ b/checkpoint-3500/rng_state_52.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:db172cb4e823361b711c304406290513980eda547fec585cfa5d76281e8cfa51 +oid sha256:ed71994ca1ece140dbf09ebf5ffbb57dcb1c3e3da83f0a91e2c2fc7a6d7962c2 size 16404 diff --git a/checkpoint-3500/rng_state_53.pth b/checkpoint-3500/rng_state_53.pth index 50aaa12600e8bff0c50f6b8d7a94ae5c431d92f6..031090b5edf297698b279dd3b548c00d0b337b86 100644 --- a/checkpoint-3500/rng_state_53.pth +++ b/checkpoint-3500/rng_state_53.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:782917ab745815b9b4a88f606bf27208a11fe179b9f93c6911f4f9c408abfea9 +oid sha256:4d466dae8b6917cc6c658a1eff20645820478cc3cd87367d087a7ebde0328359 size 16404 diff --git a/checkpoint-3500/rng_state_54.pth b/checkpoint-3500/rng_state_54.pth index 559561d4ea93c5ad8874a4674adb5eb8a5a7262f..660fdae8253f821073d898adbc2f4ea2420ac99c 100644 --- a/checkpoint-3500/rng_state_54.pth +++ b/checkpoint-3500/rng_state_54.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ab7a5e2c50dec9851e65c5aa56f5421a205f61cc179554c1f51f76449911f8a2 +oid sha256:f89c29abd7659fd7ba975f27dbaa6796a1c647363b068d139e72d80205b6de22 size 16404 diff --git a/checkpoint-3500/rng_state_55.pth b/checkpoint-3500/rng_state_55.pth index ea9684cbdbdb3d640d80a6ab21026c108852c774..43741c77bf37559eaa6ab8f5bae9c071cc2d7be0 100644 --- a/checkpoint-3500/rng_state_55.pth +++ b/checkpoint-3500/rng_state_55.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:a566a473f92a4a91bf3f90c72f39ef74967a4b72052da4ac12558a686406e136 +oid sha256:1ec775cbe40131de9c846e5394ba6311fcbb0214ab40b264805926615858b6a5 size 16404 diff --git a/checkpoint-3500/rng_state_56.pth b/checkpoint-3500/rng_state_56.pth index ab93259be7d34ad4f4d1487e78cf6ef1a3bd4925..67e5744c992a4cedffa3e3cf577541eb5a9d37fb 100644 --- a/checkpoint-3500/rng_state_56.pth +++ b/checkpoint-3500/rng_state_56.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ace4edc00ca87cc14ffa6a8fcb9c6b22548ff97189f90afc7c22814582f92ae0 +oid sha256:aa1c716333a33945f34899c5e0881325e06498d1c2fd06a6a9bb17219be1890b size 16404 diff --git a/checkpoint-3500/rng_state_57.pth b/checkpoint-3500/rng_state_57.pth index 8a360875250bd645817de66f4b434428d099ce0f..197c3903c591186cd812b6d368e8e60d709d3dd0 100644 --- a/checkpoint-3500/rng_state_57.pth +++ b/checkpoint-3500/rng_state_57.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2d635bda9ca95b0cca50b8f81b4266cfff89ac1d1907870097e8aa996a2ba1c2 +oid sha256:430e47a43f42f8eb3a29e70a0d7bab12a7697f7b592ed3b6970e44dc474d0a28 size 16404 diff --git a/checkpoint-3500/rng_state_58.pth b/checkpoint-3500/rng_state_58.pth index 0b88c9396b0a7ad49d5e2f3488ff36188d6c718c..c017db4665b96a016df10938b62bef30bf528741 100644 --- a/checkpoint-3500/rng_state_58.pth +++ b/checkpoint-3500/rng_state_58.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:8880c68f7e18dfa2c1215bc0aa962e21f041bb096c012a91ea339b89a5bd9eaa +oid sha256:b1c992ef55a8379bcad95354708defb1e991bfb4f74ed0bdb8c52998e2cf006d size 16404 diff --git a/checkpoint-3500/rng_state_59.pth b/checkpoint-3500/rng_state_59.pth index f303698a1dc5d2d87f1fd907b17cabf0059b51b6..b6fa558fe1297921406d6ed3eb4f405ae438e948 100644 --- a/checkpoint-3500/rng_state_59.pth +++ b/checkpoint-3500/rng_state_59.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:96ba53d4e66f3f146eea50ef6b58aca35e4646b6913849065944ad168a122fd8 +oid sha256:27ed28317f8060ae7271c37924b8eb021f1fb0979b0f0e4c540c07fa28a274be size 16404 diff --git a/checkpoint-3500/rng_state_6.pth b/checkpoint-3500/rng_state_6.pth index 8805f001d2c634aea1c817dc6206cd925c38b737..5ad700a49a1b1f26ac3d8160fcd5fef81063bff2 100644 --- a/checkpoint-3500/rng_state_6.pth +++ b/checkpoint-3500/rng_state_6.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f45726ac72ccb19e7c2aa4fb61800214044e82a73f7cb75adb2317929ff4f3b3 +oid sha256:7a41b56a54578445d937073710643d48256790a16367ffda22440ba03742b619 size 16389 diff --git a/checkpoint-3500/rng_state_60.pth b/checkpoint-3500/rng_state_60.pth index 49b03a87d5bf97ad7c2a51b798fe6e08050a6f11..ad8c3afa440ea780de30239234eac7bc359527e8 100644 --- a/checkpoint-3500/rng_state_60.pth +++ b/checkpoint-3500/rng_state_60.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:a79e807a7773f96ae9903ac441ead2bf0a32be441d711704f556c57e0e77d640 +oid sha256:5d9e813a00b0a367a38138b837c0ce2e9ed5022d83e4fe186df0586f60cc336d size 16404 diff --git a/checkpoint-3500/rng_state_61.pth b/checkpoint-3500/rng_state_61.pth index 35c001ec96ae6ecfb465de3eedb9206466fdbbaf..6db38440e42d0cb7aafe30a1d8b44221645e9f7e 100644 --- a/checkpoint-3500/rng_state_61.pth +++ b/checkpoint-3500/rng_state_61.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:94d93f07537639fd1fe86c815bd79ccb52a94d8bcb65f3d99c09822dd02206f5 +oid sha256:3efcf2af7032ae1d26040e9e033fdb016f286094ab0e3dbecf06a6ff4a35e058 size 16404 diff --git a/checkpoint-3500/rng_state_62.pth b/checkpoint-3500/rng_state_62.pth index 751ce34fc528fb0ee25b9f940000bf1758485dec..d47f84809d722b45750997c96d6d335b70d11f0e 100644 --- a/checkpoint-3500/rng_state_62.pth +++ b/checkpoint-3500/rng_state_62.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:4470fb937ceb01bc73ecaf68d3f7d21131ee19e856078c3a8161e7707b11e61b +oid sha256:f7162557de10b9662c7a0486e4ae7d274ef2f8c083c042da13dc17c37e1db281 size 16404 diff --git a/checkpoint-3500/rng_state_63.pth b/checkpoint-3500/rng_state_63.pth index 81099b65556aae1a4d54131ddc86977893f7d26d..d54e90b7e382c1c38a8d7e5afac6aa6c804454fc 100644 --- a/checkpoint-3500/rng_state_63.pth +++ b/checkpoint-3500/rng_state_63.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:6d31c67098d5a18b090a8d3ddaea4efed4f2e1281bcd503ce9ead17483bd1119 +oid sha256:31c4381639e1015c435823728755b4551f6b3074e62cb990ce1ad71800ec2eec size 16404 diff --git a/checkpoint-3500/rng_state_7.pth b/checkpoint-3500/rng_state_7.pth index 4601fbc994cc6e0f1206b6633313e1639e6e235c..366b7536aec7d3c68c6749dbbc0310fe890e60d7 100644 --- a/checkpoint-3500/rng_state_7.pth +++ b/checkpoint-3500/rng_state_7.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2963f6fedac4fe9b4a8d1f6956d8fb5398648907d3700257149205b486bef52f +oid sha256:00c37cad62844dbc3a7a61f0b37573cd276c2c1bc093f39bc29f70f6d647604b size 16389 diff --git a/checkpoint-3500/rng_state_8.pth b/checkpoint-3500/rng_state_8.pth index 70ff99ee0408dcc9c7b426470927476aad5ca757..1b1538eba08f7d8dd68ce0424f0431f176c73b1b 100644 --- a/checkpoint-3500/rng_state_8.pth +++ b/checkpoint-3500/rng_state_8.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f9985d0997eaef567f16b9bdff7af2211fd3bbfc15dcc9544b57276c24afb9cd +oid sha256:2673c598b04efac8e0d6584f30f6cf025a59d5f51210a6c811c801c953fa7e2c size 16389 diff --git a/checkpoint-3500/rng_state_9.pth b/checkpoint-3500/rng_state_9.pth index 2510643f5bb84e7082d77b09d36d3e9deff0d721..993bb7960e133532d45dc9b7d738f08a2715b18c 100644 --- a/checkpoint-3500/rng_state_9.pth +++ b/checkpoint-3500/rng_state_9.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:249e234a418c1a8ccb47c3569794464cc02046263d523e16ba36209d939bfed5 +oid sha256:3e27e6c0e04dfb1b1077e6a34cedd3d7128119c0d9203d477e66ce1c18514288 size 16389 diff --git a/checkpoint-3500/scheduler.pt b/checkpoint-3500/scheduler.pt index a9dad9437cbe657794610da03c17cbfe63dd514b..02aa9c3d4d4fc1f6177400a3cbfb9952e5e3542c 100644 --- a/checkpoint-3500/scheduler.pt +++ b/checkpoint-3500/scheduler.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:6261aabba0f1352071a4890693497d32be5a2bf79b697fb4127f314e8439205b +oid sha256:8f4468342936856bce0ec299eef7b86eec91d39d36ddf4ec2d1912eadbfb3a8f size 1465 diff --git a/checkpoint-3500/trainer_state.json b/checkpoint-3500/trainer_state.json index c8a524e779324a2a8aeb962bef6006b1de40ee82..81270f5ff55f934963e695020c02e9499111c036 100644 --- a/checkpoint-3500/trainer_state.json +++ b/checkpoint-3500/trainer_state.json @@ -4,7 +4,7 @@ "best_model_checkpoint": null, "epoch": 0.3744266591781335, "eval_steps": 500, - "global_step": 2000, + "global_step": 3500, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, @@ -2008,6 +2008,1446 @@ "learning_rate": 4.883275233334108e-05, "loss": 0.7164, "step": 2000 + }, + { + "diffusion_loss": 1.7221301794052124, + "epoch": 0.2583543948329121, + "grad_norm": 0.8780825059865678, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0036773681640625, + "learning_rate": 4.958942863386174e-05, + "loss": 0.7238, + "step": 2010 + }, + { + "diffusion_loss": 0.00010300547728547826, + "epoch": 0.26022652812880276, + "grad_norm": 0.6796512392091122, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00165557861328125, + "learning_rate": 4.958025940766539e-05, + "loss": 0.7559, + "step": 2020 + }, + { + "diffusion_loss": 0.2364329695701599, + "epoch": 0.26209866142469346, + "grad_norm": 0.839381471437194, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0062255859375, + "learning_rate": 4.9570989788197544e-05, + "loss": 0.7415, + "step": 2030 + }, + { + "diffusion_loss": 0.33610019087791443, + "epoch": 0.2639707947205841, + "grad_norm": 0.7549023198221578, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.001953125, + "learning_rate": 4.956161981331816e-05, + "loss": 0.718, + "step": 2040 + }, + { + "diffusion_loss": 0.7993960380554199, + "epoch": 0.2658429280164748, + "grad_norm": 0.9606654837594721, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.002471923828125, + "learning_rate": 4.955214952129707e-05, + "loss": 0.7375, + "step": 2050 + }, + { + "diffusion_loss": 0.9572303295135498, + "epoch": 0.26771506131236544, + "grad_norm": 0.8869781328188083, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0078125, + "learning_rate": 4.954257895081384e-05, + "loss": 0.7405, + "step": 2060 + }, + { + "diffusion_loss": 0.47037115693092346, + "epoch": 0.2695871946082561, + "grad_norm": 0.6949458776835219, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0037841796875, + "learning_rate": 4.953290814095758e-05, + "loss": 0.7424, + "step": 2070 + }, + { + "diffusion_loss": 0.47621989250183105, + "epoch": 0.2714593279041468, + "grad_norm": 0.8316896364450832, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0038299560546875, + "learning_rate": 4.952313713122684e-05, + "loss": 0.7463, + "step": 2080 + }, + { + "diffusion_loss": 0.07881411910057068, + "epoch": 0.27333146120003743, + "grad_norm": 0.69503443446772, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0003261566162109375, + "learning_rate": 4.951326596152939e-05, + "loss": 0.7548, + "step": 2090 + }, + { + "diffusion_loss": 0.0, + "epoch": 0.2752035944959281, + "grad_norm": 0.7754230300127414, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.002716064453125, + "learning_rate": 4.950329467218211e-05, + "loss": 0.7136, + "step": 2100 + }, + { + "diffusion_loss": 0.2782937288284302, + "epoch": 0.27707572779181877, + "grad_norm": 0.8225379967404977, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00045013427734375, + "learning_rate": 4.949322330391077e-05, + "loss": 0.7396, + "step": 2110 + }, + { + "diffusion_loss": 0.8532346487045288, + "epoch": 0.27894786108770947, + "grad_norm": 2.7404351910221063, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.04638671875, + "learning_rate": 4.9483051897849916e-05, + "loss": 0.7398, + "step": 2120 + }, + { + "diffusion_loss": 0.005874613765627146, + "epoch": 0.2808199943836001, + "grad_norm": 0.8553110902626483, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0113525390625, + "learning_rate": 4.947278049554267e-05, + "loss": 0.735, + "step": 2130 + }, + { + "diffusion_loss": 0.17747539281845093, + "epoch": 0.28269212767949076, + "grad_norm": 0.8915994450237633, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0003108978271484375, + "learning_rate": 4.946240913894058e-05, + "loss": 0.7355, + "step": 2140 + }, + { + "diffusion_loss": 0.41735342144966125, + "epoch": 0.28456426097538146, + "grad_norm": 0.9277205150864798, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.007659912109375, + "learning_rate": 4.94519378704034e-05, + "loss": 0.7116, + "step": 2150 + }, + { + "diffusion_loss": 2.929142475128174, + "epoch": 0.2864363942712721, + "grad_norm": 1.1463402773981446, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00213623046875, + "learning_rate": 4.944136673269901e-05, + "loss": 0.7327, + "step": 2160 + }, + { + "diffusion_loss": 0.43869704008102417, + "epoch": 0.2883085275671628, + "grad_norm": 0.7059496150415211, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00042724609375, + "learning_rate": 4.943069576900314e-05, + "loss": 0.7466, + "step": 2170 + }, + { + "diffusion_loss": 0.0001257308613276109, + "epoch": 0.29018066086305344, + "grad_norm": 0.689191209725196, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0018768310546875, + "learning_rate": 4.941992502289927e-05, + "loss": 0.7415, + "step": 2180 + }, + { + "diffusion_loss": 0.20411455631256104, + "epoch": 0.29205279415894414, + "grad_norm": 1.0522328821058398, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0028228759765625, + "learning_rate": 4.9409054538378405e-05, + "loss": 0.7356, + "step": 2190 + }, + { + "diffusion_loss": 0.007721802219748497, + "epoch": 0.2939249274548348, + "grad_norm": 1.0353279041315244, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00823974609375, + "learning_rate": 4.9398084359838915e-05, + "loss": 0.7409, + "step": 2200 + }, + { + "diffusion_loss": 0.2592378258705139, + "epoch": 0.2957970607507254, + "grad_norm": 0.8182814883569913, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0015869140625, + "learning_rate": 4.938701453208636e-05, + "loss": 0.721, + "step": 2210 + }, + { + "diffusion_loss": 2.5279297828674316, + "epoch": 0.2976691940466161, + "grad_norm": 1.031804839797587, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.05322265625, + "learning_rate": 4.937584510033329e-05, + "loss": 0.7444, + "step": 2220 + }, + { + "diffusion_loss": 1.837208867073059, + "epoch": 0.29954132734250677, + "grad_norm": 1.8110488568804328, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000896453857421875, + "learning_rate": 4.9364576110199054e-05, + "loss": 0.7442, + "step": 2230 + }, + { + "diffusion_loss": 1.2421565055847168, + "epoch": 0.30141346063839747, + "grad_norm": 0.7971271126813868, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0018157958984375, + "learning_rate": 4.935320760770966e-05, + "loss": 0.7114, + "step": 1610 + }, + { + "diffusion_loss": 2.011812686920166, + "epoch": 0.3032855939342881, + "grad_norm": 0.7493677677896102, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.01043701171875, + "learning_rate": 4.934173963929753e-05, + "loss": 0.7249, + "step": 2240 + }, + { + "diffusion_loss": 0.00014445446140598506, + "epoch": 0.3051577272301788, + "grad_norm": 1.2207380899007845, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000667572021484375, + "learning_rate": 4.933017225180134e-05, + "loss": 0.7327, + "step": 2250 + }, + { + "diffusion_loss": 1.8615789413452148, + "epoch": 0.30702986052606945, + "grad_norm": 0.685438968092371, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001556396484375, + "learning_rate": 4.9318505492465836e-05, + "loss": 0.7572, + "step": 2260 + }, + { + "diffusion_loss": 0.3669389486312866, + "epoch": 0.3089019938219601, + "grad_norm": 0.8942925544316402, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0113525390625, + "learning_rate": 4.9306739408941605e-05, + "loss": 0.6966, + "step": 2270 + }, + { + "diffusion_loss": 1.093544363975525, + "epoch": 0.3107741271178508, + "grad_norm": 1.1826559401973311, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.003692626953125, + "learning_rate": 4.9294874049284934e-05, + "loss": 0.7387, + "step": 2280 + }, + { + "diffusion_loss": 0.9004691243171692, + "epoch": 0.31264626041374144, + "grad_norm": 0.7491542173274212, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0247802734375, + "learning_rate": 4.928290946195756e-05, + "loss": 0.761, + "step": 2290 + }, + { + "diffusion_loss": 0.21256904304027557, + "epoch": 0.31451839370963214, + "grad_norm": 1.0690396128716897, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.4296875, + "learning_rate": 4.92708456958265e-05, + "loss": 0.7363, + "step": 2300 + }, + { + "diffusion_loss": 0.5771797895431519, + "epoch": 0.3163905270055228, + "grad_norm": 0.9040201381815194, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000705718994140625, + "learning_rate": 4.925868280016386e-05, + "loss": 0.7144, + "step": 2310 + }, + { + "diffusion_loss": 1.855783462524414, + "epoch": 0.3182626603014135, + "grad_norm": 2.2441256884776193, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001556396484375, + "learning_rate": 4.924642082464661e-05, + "loss": 0.7428, + "step": 2320 + }, + { + "diffusion_loss": 0.00025519143673591316, + "epoch": 0.3201347935973041, + "grad_norm": 0.8289843977026529, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00244140625, + "learning_rate": 4.9234059819356394e-05, + "loss": 0.7222, + "step": 2330 + }, + { + "diffusion_loss": 1.4206185340881348, + "epoch": 0.3220069268931948, + "grad_norm": 0.8978585110213869, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00058746337890625, + "learning_rate": 4.922159983477933e-05, + "loss": 0.71, + "step": 2340 + }, + { + "diffusion_loss": 0.02003294974565506, + "epoch": 0.32387906018908547, + "grad_norm": 0.978391555870651, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.205078125, + "learning_rate": 4.920904092180578e-05, + "loss": 0.709, + "step": 2350 + }, + { + "diffusion_loss": 1.0875109434127808, + "epoch": 0.3257511934849761, + "grad_norm": 0.7661004559836558, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00148773193359375, + "learning_rate": 4.9196383131730194e-05, + "loss": 0.7125, + "step": 2360 + }, + { + "diffusion_loss": 0.5352952480316162, + "epoch": 0.3276233267808668, + "grad_norm": 0.7862947833694415, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0018463134765625, + "learning_rate": 4.918362651625083e-05, + "loss": 0.7062, + "step": 2370 + }, + { + "diffusion_loss": 4.766830898006447e-06, + "epoch": 0.32949546007675745, + "grad_norm": 1.0830663257885906, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00180816650390625, + "learning_rate": 4.9170771127469596e-05, + "loss": 0.7149, + "step": 2380 + }, + { + "diffusion_loss": 0.0020996935199946165, + "epoch": 0.33136759337264815, + "grad_norm": 1.3089887314239763, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000644683837890625, + "learning_rate": 4.9157817017891815e-05, + "loss": 0.7319, + "step": 2390 + }, + { + "diffusion_loss": 0.1782829910516739, + "epoch": 0.3332397266685388, + "grad_norm": 1.2888221408189688, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00604248046875, + "learning_rate": 4.9144764240426025e-05, + "loss": 0.7404, + "step": 2400 + }, + { + "diffusion_loss": 0.02764366939663887, + "epoch": 0.3351118599644295, + "grad_norm": 1.686184736312294, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001312255859375, + "learning_rate": 4.913161284838375e-05, + "loss": 0.7184, + "step": 2410 + }, + { + "diffusion_loss": 0.20818330347537994, + "epoch": 0.33698399326032014, + "grad_norm": 2.456850960230163, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00118255615234375, + "learning_rate": 4.911836289547928e-05, + "loss": 0.7301, + "step": 2420 + }, + { + "diffusion_loss": 1.380350112915039, + "epoch": 0.3388561265562108, + "grad_norm": 0.9818947068108966, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.24609375, + "learning_rate": 4.9105014435829456e-05, + "loss": 0.7521, + "step": 2430 + }, + { + "diffusion_loss": 0.9759786128997803, + "epoch": 0.3407282598521015, + "grad_norm": 0.8446008205355949, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0042724609375, + "learning_rate": 4.909156752395345e-05, + "loss": 0.7403, + "step": 2440 + }, + { + "diffusion_loss": 1.063934326171875, + "epoch": 0.3426003931479921, + "grad_norm": 0.7143484022639192, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.02099609375, + "learning_rate": 4.9078022214772577e-05, + "loss": 0.739, + "step": 2450 + }, + { + "diffusion_loss": 0.29646003246307373, + "epoch": 0.3444725264438828, + "grad_norm": 0.9162192680085588, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0002803802490234375, + "learning_rate": 4.906437856360998e-05, + "loss": 0.7183, + "step": 2460 + }, + { + "diffusion_loss": 0.8531908988952637, + "epoch": 0.34634465973977346, + "grad_norm": 0.8833462253946767, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.01409912109375, + "learning_rate": 4.9050636626190505e-05, + "loss": 0.7145, + "step": 2470 + }, + { + "diffusion_loss": 1.5317065715789795, + "epoch": 0.34821679303566416, + "grad_norm": 0.859337078901678, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00537109375, + "learning_rate": 4.90367964586404e-05, + "loss": 0.7372, + "step": 2480 + }, + { + "diffusion_loss": 0.14512565732002258, + "epoch": 0.3500889263315548, + "grad_norm": 0.9351018195772979, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0014190673828125, + "learning_rate": 4.9022858117487135e-05, + "loss": 0.715, + "step": 2490 + }, + { + "diffusion_loss": 1.0199575424194336, + "epoch": 0.35196105962744545, + "grad_norm": 0.8594098657589293, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001617431640625, + "learning_rate": 4.900882165965915e-05, + "loss": 0.7074, + "step": 2500 + }, + { + "diffusion_loss": 0.23421645164489746, + "epoch": 0.35383319292333615, + "grad_norm": 0.8512387416064483, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000476837158203125, + "learning_rate": 4.8994687142485604e-05, + "loss": 0.7045, + "step": 2600 + }, + { + "diffusion_loss": 0.34927433729171753, + "epoch": 0.3557053262192268, + "grad_norm": 0.9717215511640022, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00119781494140625, + "learning_rate": 4.898045462369619e-05, + "loss": 0.7047, + "step": 2610 + }, + { + "diffusion_loss": 0.22166481614112854, + "epoch": 0.3575774595151175, + "grad_norm": 0.6859012448444752, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001617431640625, + "learning_rate": 4.8966124161420835e-05, + "loss": 0.7256, + "step": 2620 + }, + { + "diffusion_loss": 1.3944989442825317, + "epoch": 0.35944959281100813, + "grad_norm": 0.846271770284148, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.05517578125, + "learning_rate": 4.895169581418952e-05, + "loss": 0.7308, + "step": 2630 + }, + { + "diffusion_loss": 0.06288710981607437, + "epoch": 0.36132172610689883, + "grad_norm": 0.7909608169946507, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.002532958984375, + "learning_rate": 4.893716964093202e-05, + "loss": 0.7258, + "step": 2640 + }, + { + "diffusion_loss": 0.3706468641757965, + "epoch": 0.3631938594027895, + "grad_norm": 1.8851173155043475, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00119781494140625, + "learning_rate": 4.892254570097765e-05, + "loss": 0.7197, + "step": 2650 + }, + { + "diffusion_loss": 0.2144605815410614, + "epoch": 0.3650659926986801, + "grad_norm": 0.7233928904573314, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.002288818359375, + "learning_rate": 4.890782405405504e-05, + "loss": 0.7234, + "step": 2660 + }, + { + "diffusion_loss": 0.9534701108932495, + "epoch": 0.3669381259945708, + "grad_norm": 0.714445774054295, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00421142578125, + "learning_rate": 4.889300476029188e-05, + "loss": 0.7039, + "step": 2670 + }, + { + "diffusion_loss": 1.2137645483016968, + "epoch": 0.36881025929046146, + "grad_norm": 1.0458027366697378, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.003326416015625, + "learning_rate": 4.887808788021468e-05, + "loss": 0.731, + "step": 2680 + }, + { + "diffusion_loss": 6.462080818891991e-06, + "epoch": 0.37068239258635216, + "grad_norm": 0.7991036847276086, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.024169921875, + "learning_rate": 4.886307347474853e-05, + "loss": 0.7104, + "step": 2690 + }, + { + "diffusion_loss": 0.27051225304603577, + "epoch": 0.3725545258822428, + "grad_norm": 0.8735566267681771, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.0009765625, + "learning_rate": 4.884796160521684e-05, + "loss": 0.71, + "step": 2700 + }, + { + "diffusion_loss": 0.2564800977706909, + "epoch": 0.3744266591781335, + "grad_norm": 0.7927807838216007, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.00081634521484375, + "learning_rate": 4.883275233334108e-05, + "loss": 0.7164, + "step": 2710 + }, + { + "diffusion_loss": 0.2364329695701599, + "epoch": 0.26209866142469346, + "grad_norm": 0.839381471437194, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0062255859375, + "learning_rate": 4.9570989788197544e-05, + "loss": 0.7415, + "step": 2720 + }, + { + "diffusion_loss": 0.33610019087791443, + "epoch": 0.2639707947205841, + "grad_norm": 0.7549023198221578, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.001953125, + "learning_rate": 4.956161981331816e-05, + "loss": 0.718, + "step": 2730 + }, + { + "diffusion_loss": 0.7993960380554199, + "epoch": 0.2658429280164748, + "grad_norm": 0.9606654837594721, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.002471923828125, + "learning_rate": 4.955214952129707e-05, + "loss": 0.7375, + "step": 2740 + }, + { + "diffusion_loss": 0.9572303295135498, + "epoch": 0.26771506131236544, + "grad_norm": 0.8869781328188083, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0078125, + "learning_rate": 4.954257895081384e-05, + "loss": 0.7405, + "step": 2750 + }, + { + "diffusion_loss": 0.47037115693092346, + "epoch": 0.2695871946082561, + "grad_norm": 0.6949458776835219, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0037841796875, + "learning_rate": 4.953290814095758e-05, + "loss": 0.7424, + "step": 2760 + }, + { + "diffusion_loss": 0.47621989250183105, + "epoch": 0.2714593279041468, + "grad_norm": 0.8316896364450832, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0038299560546875, + "learning_rate": 4.952313713122684e-05, + "loss": 0.7463, + "step": 2770 + }, + { + "diffusion_loss": 0.07881411910057068, + "epoch": 0.27333146120003743, + "grad_norm": 0.69503443446772, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0003261566162109375, + "learning_rate": 4.951326596152939e-05, + "loss": 0.7548, + "step": 2780 + }, + { + "diffusion_loss": 0.0, + "epoch": 0.2752035944959281, + "grad_norm": 0.7754230300127414, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.002716064453125, + "learning_rate": 4.950329467218211e-05, + "loss": 0.7136, + "step": 2790 + }, + { + "diffusion_loss": 0.2782937288284302, + "epoch": 0.27707572779181877, + "grad_norm": 0.8225379967404977, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00045013427734375, + "learning_rate": 4.949322330391077e-05, + "loss": 0.7396, + "step": 2800 + }, + { + "diffusion_loss": 0.8532346487045288, + "epoch": 0.27894786108770947, + "grad_norm": 2.7404351910221063, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.04638671875, + "learning_rate": 4.9483051897849916e-05, + "loss": 0.7398, + "step": 2810 + }, + { + "diffusion_loss": 0.005874613765627146, + "epoch": 0.2808199943836001, + "grad_norm": 0.8553110902626483, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0113525390625, + "learning_rate": 4.947278049554267e-05, + "loss": 0.735, + "step": 2820 + }, + { + "diffusion_loss": 0.17747539281845093, + "epoch": 0.28269212767949076, + "grad_norm": 0.8915994450237633, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0003108978271484375, + "learning_rate": 4.946240913894058e-05, + "loss": 0.7355, + "step": 2830 + }, + { + "diffusion_loss": 0.41735342144966125, + "epoch": 0.28456426097538146, + "grad_norm": 0.9277205150864798, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.007659912109375, + "learning_rate": 4.94519378704034e-05, + "loss": 0.7116, + "step": 2840 + }, + { + "diffusion_loss": 2.929142475128174, + "epoch": 0.2864363942712721, + "grad_norm": 1.1463402773981446, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00213623046875, + "learning_rate": 4.944136673269901e-05, + "loss": 0.7327, + "step": 2850 + }, + { + "diffusion_loss": 0.43869704008102417, + "epoch": 0.2883085275671628, + "grad_norm": 0.7059496150415211, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00042724609375, + "learning_rate": 4.943069576900314e-05, + "loss": 0.7466, + "step": 2860 + }, + { + "diffusion_loss": 0.0001257308613276109, + "epoch": 0.29018066086305344, + "grad_norm": 0.689191209725196, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0018768310546875, + "learning_rate": 4.941992502289927e-05, + "loss": 0.7415, + "step": 2870 + }, + { + "diffusion_loss": 0.20411455631256104, + "epoch": 0.29205279415894414, + "grad_norm": 1.0522328821058398, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0028228759765625, + "learning_rate": 4.9409054538378405e-05, + "loss": 0.7356, + "step": 2880 + }, + { + "diffusion_loss": 0.007721802219748497, + "epoch": 0.2939249274548348, + "grad_norm": 1.0353279041315244, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00823974609375, + "learning_rate": 4.9398084359838915e-05, + "loss": 0.7409, + "step": 2890 + }, + { + "diffusion_loss": 0.2592378258705139, + "epoch": 0.2957970607507254, + "grad_norm": 0.8182814883569913, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0015869140625, + "learning_rate": 4.938701453208636e-05, + "loss": 0.721, + "step": 2900 + }, + { + "diffusion_loss": 2.5279297828674316, + "epoch": 0.2976691940466161, + "grad_norm": 1.031804839797587, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.05322265625, + "learning_rate": 4.937584510033329e-05, + "loss": 0.7444, + "step": 2910 + }, + { + "diffusion_loss": 1.837208867073059, + "epoch": 0.29954132734250677, + "grad_norm": 1.8110488568804328, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000896453857421875, + "learning_rate": 4.9364576110199054e-05, + "loss": 0.7442, + "step": 2920 + }, + { + "diffusion_loss": 1.2421565055847168, + "epoch": 0.30141346063839747, + "grad_norm": 0.7971271126813868, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0018157958984375, + "learning_rate": 4.935320760770966e-05, + "loss": 0.7114, + "step": 1610 + }, + { + "diffusion_loss": 2.011812686920166, + "epoch": 0.3032855939342881, + "grad_norm": 0.7493677677896102, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.01043701171875, + "learning_rate": 4.934173963929753e-05, + "loss": 0.7249, + "step": 2930 + }, + { + "diffusion_loss": 0.00014445446140598506, + "epoch": 0.3051577272301788, + "grad_norm": 1.2207380899007845, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000667572021484375, + "learning_rate": 4.933017225180134e-05, + "loss": 0.7327, + "step": 2940 + }, + { + "diffusion_loss": 1.8615789413452148, + "epoch": 0.30702986052606945, + "grad_norm": 0.685438968092371, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001556396484375, + "learning_rate": 4.9318505492465836e-05, + "loss": 0.7572, + "step": 2950 + }, + { + "diffusion_loss": 0.3669389486312866, + "epoch": 0.3089019938219601, + "grad_norm": 0.8942925544316402, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0113525390625, + "learning_rate": 4.9306739408941605e-05, + "loss": 0.6966, + "step": 2960 + }, + { + "diffusion_loss": 1.093544363975525, + "epoch": 0.3107741271178508, + "grad_norm": 1.1826559401973311, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.003692626953125, + "learning_rate": 4.9294874049284934e-05, + "loss": 0.7387, + "step": 2970 + }, + { + "diffusion_loss": 0.9004691243171692, + "epoch": 0.31264626041374144, + "grad_norm": 0.7491542173274212, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0247802734375, + "learning_rate": 4.928290946195756e-05, + "loss": 0.761, + "step": 2980 + }, + { + "diffusion_loss": 0.21256904304027557, + "epoch": 0.31451839370963214, + "grad_norm": 1.0690396128716897, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.4296875, + "learning_rate": 4.92708456958265e-05, + "loss": 0.7363, + "step": 2990 + }, + { + "diffusion_loss": 0.5771797895431519, + "epoch": 0.3163905270055228, + "grad_norm": 0.9040201381815194, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000705718994140625, + "learning_rate": 4.925868280016386e-05, + "loss": 0.7144, + "step": 3000 + }, + { + "diffusion_loss": 1.855783462524414, + "epoch": 0.3182626603014135, + "grad_norm": 2.2441256884776193, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001556396484375, + "learning_rate": 4.924642082464661e-05, + "loss": 0.7428, + "step": 3010 + }, + { + "diffusion_loss": 0.00025519143673591316, + "epoch": 0.3201347935973041, + "grad_norm": 0.8289843977026529, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00244140625, + "learning_rate": 4.9234059819356394e-05, + "loss": 0.7222, + "step": 3020 + }, + { + "diffusion_loss": 1.4206185340881348, + "epoch": 0.3220069268931948, + "grad_norm": 0.8978585110213869, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00058746337890625, + "learning_rate": 4.922159983477933e-05, + "loss": 0.71, + "step": 3030 + }, + { + "diffusion_loss": 0.02003294974565506, + "epoch": 0.32387906018908547, + "grad_norm": 0.978391555870651, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.205078125, + "learning_rate": 4.920904092180578e-05, + "loss": 0.709, + "step": 3040 + }, + { + "diffusion_loss": 1.0875109434127808, + "epoch": 0.3257511934849761, + "grad_norm": 0.7661004559836558, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00148773193359375, + "learning_rate": 4.9196383131730194e-05, + "loss": 0.7125, + "step": 3050 + }, + { + "diffusion_loss": 0.5352952480316162, + "epoch": 0.3276233267808668, + "grad_norm": 0.7862947833694415, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0018463134765625, + "learning_rate": 4.918362651625083e-05, + "loss": 0.7062, + "step": 3060 + }, + { + "diffusion_loss": 4.766830898006447e-06, + "epoch": 0.32949546007675745, + "grad_norm": 1.0830663257885906, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00180816650390625, + "learning_rate": 4.9170771127469596e-05, + "loss": 0.7149, + "step": 3070 + }, + { + "diffusion_loss": 0.0020996935199946165, + "epoch": 0.33136759337264815, + "grad_norm": 1.3089887314239763, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000644683837890625, + "learning_rate": 4.9157817017891815e-05, + "loss": 0.7319, + "step": 3080 + }, + { + "diffusion_loss": 0.1782829910516739, + "epoch": 0.3332397266685388, + "grad_norm": 1.2888221408189688, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00604248046875, + "learning_rate": 4.9144764240426025e-05, + "loss": 0.7404, + "step": 3090 + }, + { + "diffusion_loss": 0.02764366939663887, + "epoch": 0.3351118599644295, + "grad_norm": 1.686184736312294, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001312255859375, + "learning_rate": 4.913161284838375e-05, + "loss": 0.7184, + "step": 3100 + }, + { + "diffusion_loss": 0.20818330347537994, + "epoch": 0.33698399326032014, + "grad_norm": 2.456850960230163, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00118255615234375, + "learning_rate": 4.911836289547928e-05, + "loss": 0.7301, + "step": 3110 + }, + { + "diffusion_loss": 1.380350112915039, + "epoch": 0.3388561265562108, + "grad_norm": 0.9818947068108966, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.24609375, + "learning_rate": 4.9105014435829456e-05, + "loss": 0.7521, + "step": 3120 + }, + { + "diffusion_loss": 0.9759786128997803, + "epoch": 0.3407282598521015, + "grad_norm": 0.8446008205355949, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0042724609375, + "learning_rate": 4.909156752395345e-05, + "loss": 0.7403, + "step": 3130 + }, + { + "diffusion_loss": 1.063934326171875, + "epoch": 0.3426003931479921, + "grad_norm": 0.7143484022639192, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.02099609375, + "learning_rate": 4.9078022214772577e-05, + "loss": 0.739, + "step": 3140 + }, + { + "diffusion_loss": 0.29646003246307373, + "epoch": 0.3444725264438828, + "grad_norm": 0.9162192680085588, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0002803802490234375, + "learning_rate": 4.906437856360998e-05, + "loss": 0.7183, + "step": 3150 + }, + { + "diffusion_loss": 0.8531908988952637, + "epoch": 0.34634465973977346, + "grad_norm": 0.8833462253946767, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.01409912109375, + "learning_rate": 4.9050636626190505e-05, + "loss": 0.7145, + "step": 3160 + }, + { + "diffusion_loss": 1.5317065715789795, + "epoch": 0.34821679303566416, + "grad_norm": 0.859337078901678, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00537109375, + "learning_rate": 4.90367964586404e-05, + "loss": 0.7372, + "step": 3170 + }, + { + "diffusion_loss": 0.14512565732002258, + "epoch": 0.3500889263315548, + "grad_norm": 0.9351018195772979, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0014190673828125, + "learning_rate": 4.9022858117487135e-05, + "loss": 0.715, + "step": 3180 + }, + { + "diffusion_loss": 1.0199575424194336, + "epoch": 0.35196105962744545, + "grad_norm": 0.8594098657589293, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001617431640625, + "learning_rate": 4.900882165965915e-05, + "loss": 0.7074, + "step": 3190 + }, + { + "diffusion_loss": 0.23421645164489746, + "epoch": 0.35383319292333615, + "grad_norm": 0.8512387416064483, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.000476837158203125, + "learning_rate": 4.8994687142485604e-05, + "loss": 0.7045, + "step": 3200 + }, + { + "diffusion_loss": 0.34927433729171753, + "epoch": 0.3557053262192268, + "grad_norm": 0.9717215511640022, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00119781494140625, + "learning_rate": 4.898045462369619e-05, + "loss": 0.7047, + "step": 3210 + }, + { + "diffusion_loss": 0.22166481614112854, + "epoch": 0.3575774595151175, + "grad_norm": 0.6859012448444752, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.001617431640625, + "learning_rate": 4.8966124161420835e-05, + "loss": 0.7256, + "step": 3220 + }, + { + "diffusion_loss": 1.3944989442825317, + "epoch": 0.35944959281100813, + "grad_norm": 0.846271770284148, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.05517578125, + "learning_rate": 4.895169581418952e-05, + "loss": 0.7308, + "step": 3230 + }, + { + "diffusion_loss": 0.06288710981607437, + "epoch": 0.36132172610689883, + "grad_norm": 0.7909608169946507, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.002532958984375, + "learning_rate": 4.893716964093202e-05, + "loss": 0.7258, + "step": 3240 + }, + { + "diffusion_loss": 0.3706468641757965, + "epoch": 0.3631938594027895, + "grad_norm": 1.8851173155043475, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00119781494140625, + "learning_rate": 4.892254570097765e-05, + "loss": 0.7197, + "step": 3250 + }, + { + "diffusion_loss": 0.2144605815410614, + "epoch": 0.3650659926986801, + "grad_norm": 0.7233928904573314, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.002288818359375, + "learning_rate": 4.890782405405504e-05, + "loss": 0.7234, + "step": 1950 + }, + { + "diffusion_loss": 0.9534701108932495, + "epoch": 0.3669381259945708, + "grad_norm": 0.714445774054295, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00421142578125, + "learning_rate": 4.889300476029188e-05, + "loss": 0.7039, + "step": 3260 + }, + { + "diffusion_loss": 1.2137645483016968, + "epoch": 0.36881025929046146, + "grad_norm": 1.0458027366697378, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.003326416015625, + "learning_rate": 4.887808788021468e-05, + "loss": 0.731, + "step": 3270 + }, + { + "diffusion_loss": 6.462080818891991e-06, + "epoch": 0.37068239258635216, + "grad_norm": 0.7991036847276086, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.024169921875, + "learning_rate": 4.886307347474853e-05, + "loss": 0.7104, + "step": 3280 + }, + { + "diffusion_loss": 0.27051225304603577, + "epoch": 0.3725545258822428, + "grad_norm": 0.8735566267681771, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.0009765625, + "learning_rate": 4.884796160521684e-05, + "loss": 0.71, + "step": 3290 + }, + { + "diffusion_loss": 0.2564800977706909, + "epoch": 0.3744266591781335, + "grad_norm": 0.7927807838216007, + "infonce_logit_scale": 2.734375, + "infonce_loss": 0.00081634521484375, + "learning_rate": 4.883275233334108e-05, + "loss": 0.7164, + "step": 3300 + }, + { + "diffusion_loss": 1.7221301794052124, + "epoch": 0.2583543948329121, + "grad_norm": 0.8780825059865678, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0036773681640625, + "learning_rate": 4.958942863386174e-05, + "loss": 0.7238, + "step": 3310 + }, + { + "diffusion_loss": 0.00010300547728547826, + "epoch": 0.26022652812880276, + "grad_norm": 0.6796512392091122, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00165557861328125, + "learning_rate": 4.958025940766539e-05, + "loss": 0.7559, + "step": 3320 + }, + { + "diffusion_loss": 0.2364329695701599, + "epoch": 0.26209866142469346, + "grad_norm": 0.839381471437194, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0062255859375, + "learning_rate": 4.9570989788197544e-05, + "loss": 0.7415, + "step": 3330 + }, + { + "diffusion_loss": 0.33610019087791443, + "epoch": 0.2639707947205841, + "grad_norm": 0.7549023198221578, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.001953125, + "learning_rate": 4.956161981331816e-05, + "loss": 0.718, + "step": 3340 + }, + { + "diffusion_loss": 0.7993960380554199, + "epoch": 0.2658429280164748, + "grad_norm": 0.9606654837594721, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.002471923828125, + "learning_rate": 4.955214952129707e-05, + "loss": 0.7375, + "step": 3350 + }, + { + "diffusion_loss": 0.9572303295135498, + "epoch": 0.26771506131236544, + "grad_norm": 0.8869781328188083, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0078125, + "learning_rate": 4.954257895081384e-05, + "loss": 0.7405, + "step": 3360 + }, + { + "diffusion_loss": 0.47037115693092346, + "epoch": 0.2695871946082561, + "grad_norm": 0.6949458776835219, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0037841796875, + "learning_rate": 4.953290814095758e-05, + "loss": 0.7424, + "step": 3370 + }, + { + "diffusion_loss": 0.47621989250183105, + "epoch": 0.2714593279041468, + "grad_norm": 0.8316896364450832, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0038299560546875, + "learning_rate": 4.952313713122684e-05, + "loss": 0.7463, + "step": 3380 + }, + { + "diffusion_loss": 0.07881411910057068, + "epoch": 0.27333146120003743, + "grad_norm": 0.69503443446772, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0003261566162109375, + "learning_rate": 4.951326596152939e-05, + "loss": 0.7548, + "step": 3390 + }, + { + "diffusion_loss": 0.0, + "epoch": 0.2752035944959281, + "grad_norm": 0.7754230300127414, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.002716064453125, + "learning_rate": 4.950329467218211e-05, + "loss": 0.7136, + "step": 3400 + }, + { + "diffusion_loss": 0.2782937288284302, + "epoch": 0.27707572779181877, + "grad_norm": 0.8225379967404977, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00045013427734375, + "learning_rate": 4.949322330391077e-05, + "loss": 0.7396, + "step":3410 + }, + { + "diffusion_loss": 0.8532346487045288, + "epoch": 0.27894786108770947, + "grad_norm": 2.7404351910221063, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.04638671875, + "learning_rate": 4.9483051897849916e-05, + "loss": 0.7398, + "step": 3420 + }, + { + "diffusion_loss": 0.005874613765627146, + "epoch": 0.2808199943836001, + "grad_norm": 0.8553110902626483, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0113525390625, + "learning_rate": 4.947278049554267e-05, + "loss": 0.735, + "step": 3430 + }, + { + "diffusion_loss": 0.17747539281845093, + "epoch": 0.28269212767949076, + "grad_norm": 0.8915994450237633, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0003108978271484375, + "learning_rate": 4.946240913894058e-05, + "loss": 0.7355, + "step": 3440 + }, + { + "diffusion_loss": 0.41735342144966125, + "epoch": 0.28456426097538146, + "grad_norm": 0.9277205150864798, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.007659912109375, + "learning_rate": 4.94519378704034e-05, + "loss": 0.7116, + "step": 3450 + }, + { + "diffusion_loss": 2.929142475128174, + "epoch": 0.2864363942712721, + "grad_norm": 1.1463402773981446, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00213623046875, + "learning_rate": 4.944136673269901e-05, + "loss": 0.7327, + "step": 3460 + }, + { + "diffusion_loss": 0.43869704008102417, + "epoch": 0.2883085275671628, + "grad_norm": 0.7059496150415211, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.00042724609375, + "learning_rate": 4.943069576900314e-05, + "loss": 0.7466, + "step": 3470 + }, + { + "diffusion_loss": 0.0001257308613276109, + "epoch": 0.29018066086305344, + "grad_norm": 0.689191209725196, + "infonce_logit_scale": 2.703125, + "infonce_loss": 0.0018768310546875, + "learning_rate": 4.941992502289927e-05, + "loss": 0.7415, + "step": 3480 + }, + { + "diffusion_loss": 0.20411455631256104, + "epoch": 0.29205279415894414, + "grad_norm": 1.0522328821058398, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.0028228759765625, + "learning_rate": 4.9409054538378405e-05, + "loss": 0.7356, + "step": 3490 + }, + { + "diffusion_loss": 0.007721802219748497, + "epoch": 0.2939249274548348, + "grad_norm": 1.0353279041315244, + "infonce_logit_scale": 2.71875, + "infonce_loss": 0.00823974609375, + "learning_rate": 4.9398084359838915e-05, + "loss": 0.7409, + "step": 3500 } ], "logging_steps": 10,