GLM4.6-FP8 REAP
#2
by
blackcat1402
- opened
GLM4.6-FP8 REAP@25%: https://huggingface.co/cerebras/GLM-4.6-REAP-268B-A32B-FP8
GLM4.6-FP8 REAP@30%: https://huggingface.co/cerebras/GLM-4.6-REAP-252B-A32B-FP8
GLM4.6-FP8 REAP@40%: https://huggingface.co/cerebras/GLM-4.6-REAP-218B-A32B-FP8
40% REAP (cerebras/GLM-4.6-REAP-218B-A32B) is queued. Other REAP increments will not be considered for the meantime, partially due to expert count mismatch which prevents them from being converted.
Thanks for your quick reply, 40% REAP is the best one for local deployment :D