Cerebellum

community

https://github.com/deucebucket/cerebellum

AI & ML interests

Deep experimentation on ablation and quanting. Goal: trim the fat on models. Question: Is all data, good data, or, is some just noise?

Recent Activity

deucebucket updated a collection 28 days ago

Gemma 4 - Cerebellum

deucebucket updated a collection 28 days ago

North-Mini-Code - Cerebellum

deucebucket updated a collection 28 days ago

Fits 16 GB cards

View all activity

updated 3 collections 28 days ago

Gemma 4 - Cerebellum

4 items • Updated 28 days ago • 2

North-Mini-Code - Cerebellum

Coding-focused Cerebellum builds (per-group HumanEval ablation). • 1 item • Updated 28 days ago

Fits 16 GB cards

Weights under ~14 GB so a 16 GB GPU runs them fully loaded with room for context. Measured footprints on each card. • 11 items • Updated 28 days ago

updated 3 collections 30 days ago

24 GB cards and CPU offload

The big ones. Includes the 122B with its measured expert-offload recipe (about 18 GB VRAM plus 34 GB RAM). • 3 items • Updated 30 days ago

Fits 16 GB cards

Weights under ~14 GB so a 16 GB GPU runs them fully loaded with room for context. Measured footprints on each card. • 11 items • Updated 28 days ago

Fits 8 GB cards

Cerebellum builds whose weights leave context headroom on 8 GB GPUs. Measured footprints are on each card. • 3 items • Updated 30 days ago

updated a Space about 1 month ago

README