32M to 1.45B, all from scratch by one person on consumer GPUs. The 1.45B cost about $555. GGUF for Nano and Brat.
Stanislav
CastIronMind
AI & ML interests
Investigating the upper bounds of higher mathematics instruction when the instructor is a great-great-grandson of a cast-iron skillet and the student is a primate.
Recent Activity
new activity about 20 hours ago
CastIronMind/Lissa-1.5B-Instruct:Lissa 1.5B: a 1.45B model trained from scratch for $555, with the full cost accounting updated a collection 1 day ago
Lissa: from-scratch ladder, 32M to 1.45B updated a collection 1 day ago
Lissa: from-scratch ladder, 32M to 1.45BOrganizations
None yet