Prime Intellect, MoE 커널 최적화 Prime Mega MoE 공개
Prime Intellect Releases Prime Mega MoE: Trainable Fused MoE Kernel, 3.11x Faster Forward Pass
- 매체
- MarkTechPost
- 게시
- 중요도
- 3/5
- 화제 지수
- —
한국어 요약
Prime Intellect가 DeepSeek의 Mega MoE 커널 학습을 가능하게 하는 Prime Mega MoE를 공개했다. Qwen3 MoE 학습 단계를 최대 34% 가속화하며 순전파 속도는 3.11배 향상됐다.
원문 발췌 · English
Prime Intellect released Prime Mega MoE, a backward pass for DeepSeek’s fused Mega MoE kernel. It makes the kernel trainable and speeds Qwen3 MoE training steps by up to 34% in prime-rl. The post Prime Intellect Releases Prime Mega MoE: Trainable Fused MoE Kernel, 3.11x Faster Forward Pass appeared first on MarkTechPost.