Alternatives and similar repositories for TransMLA:

Users that are interested in TransMLA are comparing it to the libraries listed below

• JT-Ushio / MHA2MLA
  Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs
  ☆149Updated last week
• OpenNLPLab / lightning-attention
  Lightning Attention-2: A Free Lunch for Handling Unlimited Sequence Lengths in Large Language Models
  ☆274Updated last month
• NVlabs / COAT
  [ICLR 2025] COAT: Compressing Optimizer States and Activation for Memory-Efficient FP8 Training
  ☆168Updated last month
• mit-han-lab / duo-attention
  [ICLR 2025] DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads
  ☆442Updated last month
• step-law / steplaw
  ☆143Updated 2 weeks ago
• fla-org / native-sparse-attention
  🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"
  ☆601Updated last week
• tensorgi / T6
  The official implementation of Tensor ProducT ATTenTion Transformer (T6)
  ☆345Updated last month
• XunhaoLai / native-sparse-attention-triton
  Efficient triton implementation of Native Sparse Attention.
  ☆127Updated this week
• fla-org / flame
  🔥 A minimal training framework for scaling FLA models
  ☆82Updated last week
• RyanLiu112 / compute-optimal-tts
  Official codebase for "Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling".
  ☆231Updated last month
• FlagAI-Open / OpenSeek
  OpenSeek aims to unite the global open source community to drive collaborative innovation in algorithms, data and systems to develop next…
  ☆127Updated last week
• microsoft / SeerAttention
  SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs
  ☆89Updated 2 weeks ago
• thu-nics / MoA
  The official implementation of the paper <MoA: Mixture of Sparse Attention for Automatic Large Language Model Compression>
  ☆122Updated 3 months ago
• Outsider565 / LoRA-GA
  ☆185Updated 5 months ago
• InternLM / OREAL
  Exploring the Limit of Outcome Reward for Learning Mathematical Reasoning
  ☆161Updated last week
• dhcode-cpp / NSA-pytorch
  DeepSeek Native Sparse Attention pytorch implementation
  ☆54Updated 3 weeks ago
• fkodom / grouped-query-attention-pytorch
  (Unofficial) PyTorch implementation of grouped-query attention (GQA) from "GQA: Training Generalized Multi-Query Transformer Models from …
  ☆159Updated 10 months ago
• mdy666 / mdy_triton
  ☆108Updated this week
• RUC-GSAI / YuLan-Mini
  A highly capable 2.4B lightweight LLM using only 1T pre-training data with all details.
  ☆166Updated last week
• ZihanWang314 / CoE
  Chain of Experts (CoE) enables communication between experts within Mixture-of-Experts (MoE) models
  ☆151Updated 2 weeks ago
• astramind-ai / Mixture-of-depths
  Unofficial implementation for the paper "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models"
  ☆154Updated 9 months ago
• SkyworkAI / Skywork-MoE
  Skywork-MoE: A Deep Dive into Training Techniques for Mixture-of-Experts Language Models
  ☆130Updated 9 months ago
• NVIDIA / Star-Attention
  Efficient LLM Inference over Long Sequences
  ☆365Updated last month
• feifeibear / long-context-attention
  USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference
  ☆458Updated last week
• FasterDecoding / SnapKV
  ☆233Updated 10 months ago
• kyegomez / Mixture-of-Depths
  Implementation of the paper: "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models"
  ☆86Updated 2 weeks ago
• October2001 / Awesome-KV-Cache-Compression
  📰 Must-read papers on KV Cache Compression (constantly updating 🤗).
  ☆355Updated this week
• cmu-l3 / l1
  L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning
  ☆162Updated last week
• hemingkx / Spec-Bench
  Spec-Bench: A Comprehensive Benchmark and Unified Evaluation Platform for Speculative Decoding (ACL 2024 Findings)
  ☆243Updated last week
• thu-ml / ReMoE
  [ICLR2025] Codebase for "ReMoE: Fully Differentiable Mixture-of-Experts with ReLU Routing", built on Megatron-LM.
  ☆65Updated 3 months ago