facebookresearch / memory
Memory layers use a trainable key-value lookup mechanism to add extra parameters to a model without increasing FLOPs. Conceptually, sparsely activated memory layers complement compute-heavy dense feed-forward layers, providing dedicated capacity to store and retrieve information cheaply.
☆307Updated 3 months ago
Alternatives and similar repositories for memory:
Users that are interested in memory are comparing it to the libraries listed below
- Repo for "LoLCATs: On Low-Rank Linearizing of Large Language Models"☆223Updated last month
- PyTorch implementation of models from the Zamba2 series.☆177Updated last month
- Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients.☆196Updated 8 months ago
- A framework to study AI models in Reasoning, Alignment, and use of Memory (RAM).☆209Updated this week
- ☆158Updated last month
- [ICML 2024] CLLMs: Consistency Large Language Models☆386Updated 4 months ago
- PyTorch building blocks for the OLMo ecosystem☆165Updated this week
- ☆173Updated 3 months ago
- Efficient LLM Inference over Long Sequences☆365Updated last month
- Normalized Transformer (nGPT)☆162Updated 4 months ago
- ☆393Updated 2 weeks ago
- 🌾 OAT: A research-friendly framework for LLM online alignment, including preference learning, reinforcement learning, etc.☆224Updated last week
- [NeurIPS 2024] Official Repository of The Mamba in the Llama: Distilling and Accelerating Hybrid Models☆205Updated 2 weeks ago
- PyTorch implementation of Infini-Transformer from "Leave No Context Behind: Efficient Infinite Context Transformers with Infini-attention…☆287Updated 10 months ago
- Code for "LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding", ACL 2024☆277Updated 3 weeks ago
- OpenCoconut implements a latent reasoning paradigm where we generate thoughts before decoding.☆168Updated 2 months ago
- Muon optimizer: +>30% sample efficiency with <3% wallclock overhead☆505Updated last week
- LongRoPE is a novel method that can extends the context window of pre-trained LLMs to an impressive 2048k tokens.☆209Updated 6 months ago
- ☆182Updated this week
- Implementation of 🥥 Coconut, Chain of Continuous Thought, in Pytorch☆160Updated 2 months ago
- [ICLR 2025] DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads☆435Updated last month
- 🚀 Efficiently (pre)training foundation models with native PyTorch features, including FSDP for training and SDPA implementation of Flash…☆232Updated 2 weeks ago
- Code for Adam-mini: Use Fewer Learning Rates To Gain More https://arxiv.org/abs/2406.16793☆398Updated 3 months ago
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆123Updated 3 months ago
- Code for exploring Based models from "Simple linear attention language models balance the recall-throughput tradeoff"☆223Updated last month