DataStates / datastates-llmLinks
LLM checkpointing for DeepSpeed/Megatron
☆21Updated 2 weeks ago
Alternatives and similar repositories for datastates-llm
Users that are interested in datastates-llm are comparing it to the libraries listed below
Sorting:
- A resilient distributed training framework☆96Updated last year
- ☆74Updated 2 weeks ago
- ☆43Updated 6 months ago
- [ICML 2024] Serving LLMs on heterogeneous decentralized clusters.☆30Updated last year
- Dynamic resources changes for multi-dimensional parallelism training☆29Updated 2 months ago
- Code for MLSys 2024 Paper "SiDA-MoE: Sparsity-Inspired Data-Aware Serving for Efficient and Scalable Large Mixture-of-Experts Models"☆21Updated last year
- [OSDI'24] Serving LLM-based Applications Efficiently with Semantic Variable☆188Updated last year
- Microsoft Collective Communication Library☆66Updated 11 months ago
- Stateful LLM Serving☆87Updated 7 months ago
- SpotServe: Serving Generative Large Language Models on Preemptible Instances☆130Updated last year
- FlexFlow Serve: Low-Latency, High-Performance LLM Serving☆63Updated last month
- A framework for generating realistic LLM serving workloads☆73Updated 3 weeks ago
- PyTorch library for cost-effective, fast and easy serving of MoE models.☆252Updated 2 weeks ago
- ☆35Updated last year
- ☆47Updated last year
- An interference-aware scheduler for fine-grained GPU sharing☆150Updated 9 months ago
- Official Repo for "SplitQuant / LLM-PQ: Resource-Efficient LLM Offline Serving on Heterogeneous GPUs via Phase-Aware Model Partition and …☆34Updated 2 months ago
- PyTorch implementation of paper "Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline".☆92Updated 2 years ago
- ☆25Updated 2 years ago
- NEO is a LLM inference engine built to save the GPU memory crisis by CPU offloading☆67Updated 4 months ago
- Chimera: bidirectional pipeline parallelism for efficiently training large-scale models.☆67Updated 7 months ago
- PipeInfer: Accelerating LLM Inference using Asynchronous Pipelined Speculation☆30Updated 11 months ago
- nnScaler: Compiling DNN models for Parallel Training☆117Updated last month
- Official repository for the paper DynaPipe: Optimizing Multi-task Training through Dynamic Pipelines☆20Updated last year
- ☆136Updated last year
- Efficient Compute-Communication Overlap for Distributed LLM Inference☆61Updated 3 weeks ago
- A lightweight design for computation-communication overlap.☆182Updated 3 weeks ago
- Scalable long-context LLM decoding that leverages sparsity—by treating the KV cache as a vector storage system.☆94Updated last month
- ☆124Updated 11 months ago
- ☆21Updated last year