One external n-gram memory table, two exits for frozen LLMs: distribution fusion (quality) and speculative drafting (speed), with cheap alignment probes that predict when each pays.
-
Updated
Oct 11, 2026 - Python
#
conditional-memory
Here are 4 public repositories matching this topic...
Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models
language-modeling cognitive-architecture metacognition memory-consolidation large-language-models llm-fine-tuning conditional-memory scalable-lookup sparse-architectures
-
Updated
Jan 14, 2026 - Python
TinyEngram — a 10-lab, test-driven course to implement DeepSeek Engram conditional memory from scratch: retrieval, gating, training, inference, and systems optimization.
education tutorial sparsity pytorch triton cs336 n-gram mixture-of-experts engram llm deepseek conditional-memory
-
Updated
Nov 3, 2026 - Python
Engram conditional-memory sparse LLM framework
python deep-learning high-performance inference moe zero-dependency n-gram engram llm scaling-law deepseek memory-augmented conditional-memory sparse-lookup
-
Updated
Oct 1, 2026 - Python
Add this topic to your repo
To associate your repository with the conditional-memory topic, visit your repo's landing page and select "manage topics."