chandan11248/deepseek-innovations-from-scratch

Reverse-engineering how DeepSeek achieved frontier LLM performance at a fraction of the cost — through hands-on PyTorch implementations of MLA, MoE, MTP, RoPE, and quantization.

13
/ 100
Experimental
No License No Package No Dependents
Maintenance 10 / 25
Adoption 0 / 25
Maturity 3 / 25
Community 0 / 25

How are scores calculated?

Stars

—

Forks

—

Language

Jupyter Notebook

License

—

Category

llm-compression-optimization

Last pushed

Feb 26, 2026

Commits (30d)

0

GitHub
Llm Compression Optimization · 44 models

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/chandan11248/deepseek-innovations-from-scratch"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.

Higher-rated alternatives

ModelTC/LightCompress

[EMNLP 2024 & AAAI 2026] A powerful toolkit for compressing large models including LLMs, VLMs,...

64

p-e-w/heretic

Fully automatic censorship removal for language models

62

Orion-zhen/abliteration

Make abliterated models with transformers, easy and fast

54

YerbaPage/LongCodeZip

LongCodeZip: Compress Long Context for Code Language Models [ASE2025]

54

locuslab/wanda

A simple and effective LLM pruning approach.

47

Explore Transformer Models

All categories Trending Transformer directory Insights