Skip to content

Implement DeepSeek-V4 Compressed Attention Layers#3866

Draft
parambole wants to merge 2 commits into
dsv4-moe-routing-primitivesfrom
deepseek_v4_compressed_attention
Draft

Implement DeepSeek-V4 Compressed Attention Layers#3866
parambole wants to merge 2 commits into
dsv4-moe-routing-primitivesfrom
deepseek_v4_compressed_attention