Kris Bailey PRO

krisbailey

AI & ML interests

quantization, optimization, novel model architectures, model architecture research and development, dataset construction, apple silicon optimizations

Recent Activity

posted an update 14 days ago

Just posted an article about my adventures with Qwen3.8-Next-Flash running on a DGX Spark. Got up to 114 tok/s across 8 concurrent streams. https://huggingface.co/blog/krisbailey/shortlist-mtp-mostly-worked-around-a-bad-kernel-ch

published an article 14 days ago

Shortlist MTP Mostly Worked Around a Bad Kernel Choice

new activity 7 months ago

krisbailey/cosmopedia-10B:[bot] Conversion to Parquet

View all activity

Organizations

None yet

posted an update 14 days ago

Post

published an article 14 days ago

Article

Shortlist MTP Mostly Worked Around a Bad Kernel Choice

krisbailey

•

14 days ago

New activity in krisbailey/cosmopedia-10B 7 months ago

[bot] Conversion to Parquet

#1 opened 8 months ago by

parquet-converter

posted an update 7 months ago

Post

520

While doing various projects I kept running into situations where I wanted to be able to have representative samples of some of the current large SOTA datasets that were smaller so I didn't need to worry about slicing or anything else at runtime. So, I created sub datasets making sure to keep the same ratios of data sources. Each dataset card provides info for what's in it.
100M token datasets:
RedPajama v2 100M
Falcon RefinedWeb 100M
Cosmopedia 100M

1B token datasets:
Fineweb-edu 1B
RedPajama v1 1B
RedPajama v2 1B (use this one)
Cosmopedia 1B

10B token datasets:
RedPajama v1 10B
Cosmopedia 10B

Collection here:
https://huggingface.co/collections/krisbailey/bite-size-data