GenomaLabs-com
/

kv-cache-eviction-mla

inference-optimization

Model card Files Files and versions

kv-cache-eviction-mla / notebooks

Commit History

B3a-pivot: real Kimi K2.6 weights demo - eviction policy validated on actual MLA attention distribution

aac6c2e

GENOMA LABS / research commited on May 8

B1 validation: multi-step eviction test + transformers compatibility note

1ba26d6

GENOMA LABS / research commited on May 8

initial release: H2O KV cache eviction for DeepseekV3 / MLA architectures

a8d4591

GENOMA LABS / research commited on May 8