NVIDIA / NVIDIA/TensorRT-LLM

MoE optimizations

Open
#3,962 0 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

Investigating Performance roadmap triaged
Dominant language
Python
Stars
14.7k
Forks
2.8k
Avg merge
2d 23h
Merged PRs (30d)
489

Description

  • General optimization
    • [Ongoing] SmartRouter(targeting min-latency)
    • [Ongoing] Large-scale EP (custom large-scale A2A + EP workload balancer)
      • First targeting GB200 NVL72
      • Extending support for EP across nodes
    • [Ongoing] Multi-shot Allreduce optimization on GB200
  • Optimizations for DeepSeek R1
    • Per-Tensor FP8 KV Cache
      • [Done] Hopper
      • [Ongoing] Blackwell
    • [Ongoing] KV Cache reuse
    • [Ongoing] Chunked context
    • [Ongoing] INT4 AWQ

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No files, tests, or entry points are named. Begin by locating the MoE, DeepSeek R1, SmartRouter, expert-parallelism, and KV-cache implementations in the repository, then check which listed optimization still has maintainer direction. Done requires a defined target, implementation scope, and validation criteria for one optimization rather than the full roadmap.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
distributed-systems, machine-learning, performance
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
15/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.