foundation-model-stack / fms-extras
β24Updated 8 months ago
Alternatives and similar repositories for fms-extras:
Users that are interested in fms-extras are comparing it to the libraries listed below
- π Efficiently (pre)training foundation models with native PyTorch features, including FSDP for training and SDPA implementation of Flashβ¦β244Updated this week
- Example of applying CUDA graphs to LLaMA-v2β12Updated last year
- A safetensors extension to efficiently store sparse quantized tensors on diskβ104Updated this week
- Simple implementation of Speculative Sampling in NumPy for GPT-2.β95Updated last year
- Triton-based implementation of Sparse Mixture of Experts.β212Updated 5 months ago
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clustersβ126Updated 5 months ago
- Flash-Muon: An Efficient Implementation of Muon Optimizerβ100Updated this week
- Repository for Sparse Finetuning of LLMs via modified version of the MosaicML llmfoundryβ40Updated last year
- DPO, but faster πβ42Updated 5 months ago
- π Collection of components for development, training, tuning, and inference of foundation models leveraging PyTorch native components.β194Updated this week
- β104Updated 8 months ago
- Demo of the unit_scaling library, showing how a model can be easily adapted to train in FP8.β45Updated 9 months ago
- A collection of reproducible inference engine benchmarksβ29Updated 2 weeks ago
- The source code of our work "Prepacking: A Simple Method for Fast Prefilling and Increased Throughput in Large Language Models" [AISTATS β¦β59Updated 7 months ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.β122Updated this week
- [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decodingβ116Updated 5 months ago
- The simplest implementation of recent Sparse Attention patterns for efficient LLM inference.β60Updated 3 months ago
- Experiments on speculative sampling with Llama modelsβ126Updated last year
- Load compute kernels from the Hubβ116Updated this week
- Official repository for DistFlashAttn: Distributed Memory-efficient Attention for Long-context LLMs Trainingβ209Updated 8 months ago
- Odysseus: Playground of LLM Sequence Parallelismβ69Updated 10 months ago
- Using FlexAttention to compute attention with different masking patternsβ43Updated 7 months ago
- β54Updated this week
- prime-rl is a codebase for decentralized RL training at scaleβ89Updated this week
- β68Updated last month
- Make triton easierβ47Updated 10 months ago
- Boosting 4-bit inference kernels with 2:4 Sparsityβ73Updated 8 months ago
- A place to store reusable transformer components of my own creation or found on the interwebsβ55Updated last week
- Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking cache compression methods built on top ofβ¦β131Updated 9 months ago
- A toolkit for fine-tuning, inferencing, and evaluating GreenBitAI's LLMs.β83Updated last month