Updated Manifest.json for UNIGRAM variants.

And some 8k shards..

Missed some shards on UNIGRAM 16k

Added 2k UNIGRAM variant.

Added 4k UNIGRAM variants.

Added 8k Unigram variant

Added 16k Unigram variants.

Manifest updated.

Added 16k vocab variant.

Added 8k vocab + files (126 + 1 val).

Fix: I forgot a comma.

Added 256nb to manifest.:

Added 256nb tokenizer + shards (256 + val).

Added 50 shard variant.

Added 75 shard compressed artifact.

Updated manifest to reflect new types.

Added 4096nb shards (0-140 + val) and compressed artifacts.

Added 2048nb shards (0-162 + val) and compressed artifacts.

Added raw sp512nb shards (0-182 + val).

Add sp1024nb compressed shards (train 0-99 + val and train 0-24 + val)

Added raw sp512nb shards (0-212 + val).

Rename datasets/datasets/fineweb10B_sp384nb/fineweb_train_first_25_shards.tar.zst to datasets/datasets/fineweb10B_sp384nb_25.tar.zst

Rename datasets/datasets/fineweb10B_sp384nb/fineweb_shards_train_first_100_shards.tar.zst to datasets/datasets/fineweb10B_sp384nb_100.tar.zst

Add sp512nb compressed shards (train 0-99 + val)

Add sp512nb compressed shards (train 0-24 + val)

Add sp512nb compressed shards (train 0-25 + val)

Add 4096 no-byte-fallback tokenizer

Rename 384 tokenizer to 384nb (no byte_fallback)

Rename 512 tokenizer to 512nb (no byte_fallback)

Add 2048 no-byte-fallback tokenizer

Added 1024nb variant (model and vocab).

Added 512 vocab model and vocab.

Added compressed bundle of the first 100 sb384nb shards (plus validation shard).

Added compressed bundle of the first 25 sb384nb shards (plus validation shard).

Added tokenizer_config.export.

Added the 384 vocab shards.

Added new 384 vocab.

Added 384 tokenizer config.

README.txt

Duplicate from willdepueoai/parameter-golf