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
12.67k stars 864 forks source link

RWKV 5 supported vLLM？LMdeploy？TGI？Fastllm？FasterTransformer？ #232

Open lanzhoushaobing opened 8 months ago

lanzhoushaobing commented 8 months ago

RWKV 5 supported vLLM？LMdeploy？TGI？Fastllm？FasterTransformer？ What should I do to get the inference performance？like throughput, token latency and latency？

BlinkDL commented 7 months ago

not yet. someone is working on rwkv.cpp support https://github.com/ggerganov/llama.cpp/issues/846

harrisonvanderbyl commented 1 month ago

https://github.com/vllm-project/vllm/pull/8787 still wip