Commit History
5 days ago
-
Measured speed in the model card and the docstring: first kernel version 1.23x faster than dense; current version 36% higher training throughput than dense (306M, d_ff 4096, 6.25% active)
44b228d
15 days ago
-
Model card summary; English only; scripts renamed (speed_check / demo_collapse)
0c0ac02 -
Initial release: code, docs, tests (CC BY-NC 4.0)
8485f65