Add files using upload-large-folder tool

README.md

@@ -26,7 +26,7 @@ here — "uncurated" describes the *absence* of any quality/dedup/balance/safety
 beyond what upstream CommonPool already does.
 
 **License / use:** non-commercial research only, consistent with the upstream DataComp /
-CommonPool terms. This repository is private to the dataset owner.
+CommonPool terms.
 
 ## Structure: data vs. manifest
 
@@ -53,15 +53,18 @@ without touching the rest of the pool**:
 
 ## Splits — defined at the shard level, not the sample level
 
-Samples are globally shuffled once, then assigned to splits in whole **shard-sized chunks**
-(default 10,000 samples/shard): the first N shards become `val`, the next M become `test`,
-everything after that is `train`. This means:
+Samples are randomized via a streaming shuffle (raw input shard order is shuffled, plus a
+50,000-sample sliding shuffle buffer downstream — chosen so the full ~5.7M-image build never
+has to hold the whole pool in memory), then assigned to splits in whole **shard-sized chunks**
+(10,000 samples/shard): the first chunk becomes `val`, the next becomes `test`, everything
+after that is `train`. This means:
 
 - Splits are **disjoint sets of tar files**, so pulling `val` or `test` (or a `train` scale
   tier) is just downloading those specific shards — no need to touch `manifest.parquet` or
   decode unrelated shards first.
-- No fixed random seed is used for the shuffle — this artifact is meant to be used as-is, not
-  regenerated bit-for-bit.
+- No fixed random seed is used — this artifact is meant to be used as-is, not regenerated
+  bit-for-bit. The shuffle is good enough to avoid gross ordering bias but is not a uniform
+  global permutation.
 
 > **Note on val/test:** this pool is unlabeled. `val`/`test` are held-out *image* sets for
 > loss monitoring / leakage control during pretraining, not a labeled evaluation set.
@@ -79,5 +82,7 @@ everything after that is `train`. This means:
 
 ## Versions
 
-- `v0.1` — pilot, ~130 raw input shards.
-- `v1.0` — full build, all raw input shards (~5.7M images pre-filter).
+- `v0.1` — pilot. 130 raw input shards -> 112,046 samples (1 val + 1 test + 10 train shards).
+- `v1.0` — full build. All 6,540 raw input shards, 5,799,314 raw samples seen ->
+  **5,684,754 kept** (1 val + 1 test + 567 train shards, ~151GB). Scale tiers:
+  `xs`=570k/57 shards, `s`=2.84M/284 shards, `m`=5.67M/567 shards (all of train).

data/test/test-000000.tar

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:70ec0f7c12a850bbef90ee61a683d32850f5a0e764354aebbcccd67da6361241
-size 282173440
+oid sha256:dd8c3258fbfe0774a63e776dcac03ed74c43ffc88245719c343c61b106f16313
+size 282368000

data/train/train-000260.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5b12ff592034454f2964e6783fea1336e92d42545f71937a022d1d0db66057d
+size 280688640

data/train/train-000261.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe08141aff976bce27d845cd00caabfae8e9122034a8675d44fb2c2b5d974665
+size 279726080

data/train/train-000262.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3baae44a1f2bafdd1f86de43b259146f5be39d21cb1cb29751809cf6791bb8a9
+size 282214400

data/train/train-000263.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86002bf27e4ca641d71894935743d988a307e914ee8a53ed4434f85131422f29
+size 283586560

data/train/train-000264.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c7efdc9fa27a42099bbc756bfcad2465ba19e5b80f1e9b960042917fd529d1b
+size 282398720

data/train/train-000265.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d55520d7a37c5218281a14042ceb1a2cd154d93e0b4191e25844a1ce93cea003
+size 281047040

data/train/train-000266.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95c9ef70751f4836f9fd016f8c5c582eb0f9631d193b2017d37a49fa4fcbd6bc
+size 280299520

data/train/train-000267.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e36c170ad8e9a207e0b526058e3443cdc90838454321bf40b8994cd23af82d93
+size 279592960

data/train/train-000268.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5e946a8159d3333df4c08cd751d5aa838d4d8a5284fe39dc058ceadc7080c70
+size 280442880

data/train/train-000269.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b352c05b5a9c1f59b7672ea2d5146f6f9b8192a5ad3d77828c8f3a6d31a8899
+size 280094720

data/train/train-000270.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0dbd59c0f2bda7c7ddf6f954f1ecf9d432ecd754f7f8a1fbae2da2f31fcc8cfb
+size 280350720

data/train/train-000271.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:487aea6b39d75573ce6c99841a40647e54acdfd9a11dbafb5adc6ee7e9bfd57d
+size 281047040

data/train/train-000272.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e99f56f3cecd429b33d225ea381e2fbaf3fe78368b5481dff9a833d7c0c1b23f
+size 281466880

data/train/train-000273.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92f5cddc4b3ec1ec088e504aa7b8d3235090235c187b63cdcda36b4c0a93e2d4
+size 279736320

data/train/train-000274.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b5b2508ff85ae711eff91ade018542fc877f40239688f9a20265027f54c2cb1
+size 281303040

data/train/train-000275.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37129ea9c7ad666d7df6a19411682d28e785c53120f91029a4990d4e4d46f243
+size 280023040

data/train/train-000276.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:697384962df8e95f5ec289ecaf29fa8c76d6efc0f7e91930070f577dbd54c62c
+size 281262080

data/train/train-000277.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e23a2240b51e82a3e5f222fdf4eed5b8a6b9494d492e14451750fab760c7a27c
+size 280995840

data/train/train-000278.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff73960a78a5a8cdba1774c6fd5e15beb105e3e60b62c7078da1cb72d12d07bd
+size 281313280

data/train/train-000279.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d040bc1f9872923284257e7a55773911038aaea54f0093fedb92c751e180cdf
+size 280903680

data/train/train-000280.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9550d1e9ab452b2e418ca5bd5638ae45024f905d7a686f83b30d1cbceaf8447f
+size 281200640

data/train/train-000281.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06288c18354e05a025538af0fa97da934a66c5ee90bf9f04152ef642be0243e5
+size 281774080

data/train/train-000282.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b046d0bb807c367b0ae8f206bc49823493d609f5478d1fcc6d4cf7df7cde12fa
+size 281180160

data/train/train-000283.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77a7c5799a821af6bcfda0a44b61496bb260be385af05cfd43911633dc415413
+size 281610240

data/train/train-000284.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2705824150608604719fadfcf1a724d1d431574e8c76899b3610da07cdb819ac
+size 280832000

data/train/train-000285.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a91ee34bb21c90b017b46c4cf8cea7657652bc34ace2f73c2767e89e7b999b6
+size 281559040

data/train/train-000286.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ec08dec2a1ebf7fbc65d7338c8a0cbdcbca4675c60e14cab0235cb32609dfd0
+size 282828800

data/train/train-000287.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec3c48cfb16f0dda6632df29f5e14f553bf1559e2512c653cff75fe772c830cd
+size 280657920

data/train/train-000288.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:feb144664d2c6ef6a049ef0d93eafaeb2a31d6ab94cb6462b0022907eecbcc0e
+size 281446400

data/train/train-000289.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a137681bcefafb7cb928048f93fb7c7f601429248036208edc096c1fb723689c
+size 282726400

data/train/train-000290.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1172c63909570cdc2340df61c9f419da84e9f3dd43fe0d46fd26f8dad082a62e
+size 281016320

data/train/train-000291.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e13d1e6e942add8823b21a9417b33afede857b6d3c2acc790842ea70992ae7a
+size 280678400

data/train/train-000292.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:346c434c64e51ece0fa49bba7b28a70e9b30fdb255e3f6ecfb6bda04b873c3b2
+size 281507840

data/train/train-000293.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e582b2f0e6159e3f32c4682ca974fd3ac2ab386d3a0fd9519b740081bb72efad
+size 279582720

data/train/train-000294.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90766a9a91065a8284e95e3fa87bf73d32aa3bfda56bfc7f2e0eb1b3b63ee74b
+size 281241600

data/train/train-000295.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2735793b8d6215e396b00597ee773b92b69204bd3a3aed36a294cb330617ee3a
+size 282419200

data/train/train-000296.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9b3bc470db2895013faabab9e72c7cf98b5d6a2031f531c54b9c99207de93d9
+size 281395200

data/train/train-000297.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a55112968c6569878414e0d0b3df4a7ffd28ed472dc9b62820e1eb3e6dce80a
+size 281282560

data/train/train-000298.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8af1c36b65fcc5066c7401d5951d52394e26d3cf9b5ef5ea88f8b2a5c583865c
+size 281804800

data/train/train-000299.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6a57769e90e2b1ee449df2e6f762cab959b5b42f1ef6a80d8d6f82c6075f583
+size 280514560

data/train/train-000300.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3356ef117cace71e6d213f002df577b080af0c34633cda411e3962d9dcf2fc6
+size 280883200

data/train/train-000301.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:441563dd4a97ec7c091c1a5cca111eff0850fa1576481518fc38274827b88096
+size 280729600

data/train/train-000302.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69ea4aa0db19c8c3346a97fb6d6193d25df43f2cac7c30ef0e08796398cf6de3
+size 280616960

data/train/train-000303.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db214fc42e7d706ea13230b34cece2b912380172087661669b6471e8ec2f6d97
+size 279808000

data/train/train-000304.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5ecd81dabeeb6024f250bb68bd5b4d87d15f29066922ae9b6d32965cfad8a71
+size 281958400

data/train/train-000305.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ebdbc7415ab7eed5a257ddab4c7cfb254a261df083952d05968da8bdef33f97
+size 282173440

data/train/train-000306.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d427f43ff5c62dbca4fff56b8d0be17166a3934cb5f1af7c86d47c8f07760387
+size 281047040

data/train/train-000307.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e6313eab15033e71dc025391410b1d540f039ebc9b2ab06cdb950a9ae69e586
+size 279162880