thuml / learn_torch.compileLinks
torch.compile artifacts for common deep learning models, can be used as a learning resource for torch.compile
☆17Updated last year
Alternatives and similar repositories for learn_torch.compile
Users that are interested in learn_torch.compile are comparing it to the libraries listed below
Sorting:
- IntLLaMA: A fast and light quantization solution for LLaMA☆18Updated last year
- 🤗CacheDiT: A Training-free and Easy-to-use Cache Acceleration Toolbox for Diffusion Transformers🔥☆61Updated this week
- Quantized Attention on GPU☆44Updated 7 months ago
- An auxiliary project analysis of the characteristics of KV in DiT Attention.☆31Updated 6 months ago
- ☆49Updated last month
- Benchmark tests supporting the TiledCUDA library.☆16Updated 7 months ago
- ☆39Updated this week
- Odysseus: Playground of LLM Sequence Parallelism☆70Updated last year
- Accelerate LLM preference tuning via prefix sharing with a single line of code☆41Updated last month
- ☆71Updated last month
- ☆21Updated last month
- Patch convolution to avoid large GPU memory usage of Conv2D☆88Updated 5 months ago
- GPTQ inference TVM kernel☆40Updated last year
- Framework to reduce autotune overhead to zero for well known deployments.☆77Updated last week
- Code for Draft Attention☆72Updated last month
- FastCache: Fast Caching for Diffusion Transformer Through Learnable Linear Approximation [Efficient ML Model]☆24Updated 3 weeks ago
- Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.☆38Updated 2 weeks ago
- A Suite for Parallel Inference of Diffusion Transformers (DiTs) on multi-GPU Clusters☆47Updated 11 months ago
- ☆26Updated last year
- ☆19Updated 9 months ago
- DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling☆17Updated 2 weeks ago
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆80Updated last month
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆90Updated 2 weeks ago
- ☆75Updated 5 months ago
- 🎬 3.7× faster video generation E2E 🖼️ 1.6× faster image generation E2E ⚡ ColumnSparseAttn 9.3× vs FlashAttn‑3 💨 ColumnSparseGEMM 2.5× …