bytedance / ABQ-LLM
☆23Updated this week
Related projects: ⓘ
- 使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention☆46Updated last month
- A standalone GEMM kernel for fp16 activation and quantized weight, extracted from FasterTransformer☆82Updated 6 months ago
- ☆67Updated last week
- [ACL 2024] A novel QAT with Self-Distillation framework to enhance ultra low-bit LLMs.☆69Updated 4 months ago
- QQQ is an innovative and hardware-optimized W4A8 quantization solution.☆59Updated last week
- Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.☆20Updated last week
- ☆102Updated 3 months ago
- Standalone Flash Attention v2 kernel without libtorch dependency☆93Updated last week
- GPTQ inference TVM kernel☆35Updated 4 months ago
- An easy-to-use package for implementing SmoothQuant for LLMs☆78Updated 4 months ago
- Summary of some awesome work for optimizing LLM inference☆26Updated this week
- ☆140Updated 4 months ago
- PyTorch bindings for CUTLASS grouped GEMM.☆41Updated 3 weeks ago
- play gemm with tvm☆81Updated last year
- llama INT4 cuda inference with AWQ☆46Updated 2 months ago
- ☆56Updated this week
- Summary of system papers/frameworks/codes/tools on training or serving large model☆56Updated 9 months ago
- ☆33Updated 5 months ago
- ☆77Updated last year
- ☆133Updated 2 months ago
- ☆14Updated last week
- Odysseus: Playground of LLM Sequence Parallelism☆50Updated 3 months ago
- High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.☆81Updated 2 months ago
- TiledCUDA is a highly efficient kernel template library designed to elevate CUDA C’s level of abstraction for processing tiles.☆114Updated last week
- ☆60Updated last month
- Penn CIS 5650 (GPU Programming and Architecture) Final Project☆21Updated 9 months ago
- ☆52Updated this week
- ☆34Updated 2 years ago
- ☆29Updated 4 months ago
- A parallelism VAE avoids OOM for high resolution image generation☆34Updated 2 months ago