fascinating
#1
by
KnutJaegersberg
- opened
what's this model? compared with first rwkv flash?
its L12 Hybrid.
first release "Gen1" uses L6 GQA + L38 RWKV
second release "Gen2" uses L12 GQA + L32 RWKV
Gen1 could 40k NIAH.
Gen2 reached 65k NIAH(more ctx len, stable reasoning, fine-tuning capability)
will write description :)