search

BlinkDL / RWKV-LM

RWKV is an RNN with transformer-level LLM performance. It can be directly trained like a GPT (parallelizable). So it's combining the best of RNN and transformer - great performance, fast inference, saves VRAM, fast training, "infinite" ctx_len, and free sentence embedding.
Apache License 2.0
11.99k stars 825 forks source link

MoE support #212

Open James4Ever0 opened 6 months ago

James4Ever0 commented 6 months ago

Mixtral 8x7b is out. Is there any plan for RWKV to support MoE in the future, with inference speedup? Looking forward to it.

BlinkDL commented 6 months ago

Yes :) Likely in a few months.