diff --git a/compilers/apex.yaml b/compilers/apex.yaml index b7ea663..a7facf9 100644 --- a/compilers/apex.yaml +++ b/compilers/apex.yaml @@ -8,6 +8,7 @@ apex: - nvidia - bsd-3-clause + temperature: hot url: https://github.com/NVIDIA/apex description: | diff --git a/compilers/bladedisc.yaml b/compilers/bladedisc.yaml index 5f3eb58..354be6e 100644 --- a/compilers/bladedisc.yaml +++ b/compilers/bladedisc.yaml @@ -8,6 +8,7 @@ bladedisc: - mlir - apache-2.0 + temperature: hot url: https://github.com/alibaba/BladeDISC description: | diff --git a/compilers/candle.yaml b/compilers/candle.yaml index abcb64f..c5910a7 100644 --- a/compilers/candle.yaml +++ b/compilers/candle.yaml @@ -10,6 +10,7 @@ candle: - mit - apache-2.0 + temperature: hot url: https://github.com/huggingface/candle description: | diff --git a/compilers/executorch.yaml b/compilers/executorch.yaml index 8f81a3e..80656ad 100644 --- a/compilers/executorch.yaml +++ b/compilers/executorch.yaml @@ -9,6 +9,7 @@ executorch: - edge - bsd-3-clause + temperature: hot url: https://pytorch.org/executorch-overview description: | diff --git a/compilers/flexgen.yaml b/compilers/flexgen.yaml index 47e955f..851ee0b 100644 --- a/compilers/flexgen.yaml +++ b/compilers/flexgen.yaml @@ -10,6 +10,7 @@ flexgen: - llm - apache-2.0 + temperature: hot url: https://github.com/FMInference/FlexGen description: | diff --git a/compilers/ggml.yaml b/compilers/ggml.yaml index 5d8e57d..768d278 100644 --- a/compilers/ggml.yaml +++ b/compilers/ggml.yaml @@ -12,6 +12,7 @@ ggml: - compression - mit + temperature: hot url: https://ggml.ai description: | diff --git a/compilers/glow.yaml b/compilers/glow.yaml index bed623f..615a2cb 100644 --- a/compilers/glow.yaml +++ b/compilers/glow.yaml @@ -7,6 +7,7 @@ glow: - compilers - apache-2.0 + temperature: cold url: https://ai.meta.com/tools/glow/ description: | diff --git a/compilers/hidet.yaml b/compilers/hidet.yaml index dca7822..b58d577 100644 --- a/compilers/hidet.yaml +++ b/compilers/hidet.yaml @@ -8,6 +8,7 @@ hidet: - pytorch - apache-2.0 + temperature: hot url: https://github.com/hidet-org/hidet description: | diff --git a/compilers/ipex.yaml b/compilers/ipex.yaml index 2eccd24..34a55d9 100644 --- a/compilers/ipex.yaml +++ b/compilers/ipex.yaml @@ -8,6 +8,7 @@ ipex: - intel - apache-2.0 + temperature: hot url: https://github.com/intel/intel-extension-for-pytorch description: | diff --git a/compilers/iree.yaml b/compilers/iree.yaml index adadcfa..db4af25 100644 --- a/compilers/iree.yaml +++ b/compilers/iree.yaml @@ -9,6 +9,7 @@ iree: - mlir - apache-2.0 + temperature: hot url: https://iree.dev description: | diff --git a/compilers/keops.yaml b/compilers/keops.yaml index f70c5b8..f29785d 100644 --- a/compilers/keops.yaml +++ b/compilers/keops.yaml @@ -7,6 +7,7 @@ keops: - compilers - mit + temperature: neutral url: https://www.kernel-operations.io/keops/index.html description: | diff --git a/compilers/kernl.yaml b/compilers/kernl.yaml index ac951c8..bef9869 100644 --- a/compilers/kernl.yaml +++ b/compilers/kernl.yaml @@ -8,6 +8,7 @@ kernl: - pytorch - apache-2.0 + temperature: neutral url: https://www.kernl.ai description: | diff --git a/compilers/mlc-llm.yaml b/compilers/mlc-llm.yaml index cd2a70b..603fc51 100644 --- a/compilers/mlc-llm.yaml +++ b/compilers/mlc-llm.yaml @@ -8,6 +8,7 @@ mlc-llm: - llm - apache-2.0 + temperature: hot url: https://llm.mlc.ai description: | diff --git a/compilers/mlgo.yaml b/compilers/mlgo.yaml index 91a90da..3410580 100644 --- a/compilers/mlgo.yaml +++ b/compilers/mlgo.yaml @@ -7,6 +7,7 @@ mlgo: - compilers - apache-2.0 + temperature: neutral url: https://github.com/google/ml-compiler-opt description: | diff --git a/compilers/mlir.yaml b/compilers/mlir.yaml index e6fc845..d9ad917 100644 --- a/compilers/mlir.yaml +++ b/compilers/mlir.yaml @@ -9,6 +9,7 @@ mlir: - mlir - apache-2.0 + temperature: hot url: https://mlir.llvm.org description: | diff --git a/compilers/mojo.yaml b/compilers/mojo.yaml index 16f3849..082d862 100644 --- a/compilers/mojo.yaml +++ b/compilers/mojo.yaml @@ -8,6 +8,7 @@ mojo: - mlir - proprietary + temperature: hot url: https://www.modular.com/mojo description: | diff --git a/compilers/oneapi.yaml b/compilers/oneapi.yaml index 988b2a1..b81ca18 100644 --- a/compilers/oneapi.yaml +++ b/compilers/oneapi.yaml @@ -9,6 +9,7 @@ oneapi: - other-license - mit + temperature: neutral url: https://www.oneapi.io description: | diff --git a/compilers/oneflow.yaml b/compilers/oneflow.yaml new file mode 100644 index 0000000..0a76c43 --- /dev/null +++ b/compilers/oneflow.yaml @@ -0,0 +1,27 @@ +oneflow: + name: "Oneflow" + + image_url: https://docs.oneflow.org/en/master/assets/product-layer.png + + tags: + - compilers + - framework + - mlir + - apache-2.0 + + temperature: hot + url: https://docs.oneflow.org/en/master/index.html + + description: | + OneFlow is a deep learning framework that offers a unified solution for both deep learning + and traditional machine learning tasks. It stands out for its efficient approach to distributed + training, leveraging advanced parallelism and resource management techniques to optimize hardware + usage in large-scale environments. The framework supports both dynamic and static computation graphs, + providing users with the flexibility to choose the most suitable approach for their specific project. + Additionally, OneFlow utilizes MLIR for its codegen, with all modules being compiled to a Oneflow dialect + before lowering to device code. + + features: + - "Advanced Distributed Training Efficiency" + - "Support for Both Dynamic and Static Graphs" + - "Intuitive and User-Friendly API" diff --git a/compilers/pi.yaml b/compilers/pi.yaml index 10ef7b4..29eceb6 100644 --- a/compilers/pi.yaml +++ b/compilers/pi.yaml @@ -9,6 +9,7 @@ pi: - pytorch - apache-2.0 + temperature: neutral url: https://github.com/nod-ai/PI#installing description: | diff --git a/compilers/plaidml.yaml b/compilers/plaidml.yaml index 09d0c21..f0dfb95 100644 --- a/compilers/plaidml.yaml +++ b/compilers/plaidml.yaml @@ -9,6 +9,7 @@ plaidml: - mlir - apache-2.0 + temperature: cold url: https://plaidml.github.io/plaidml/ description: | diff --git a/compilers/polyblocks.yaml b/compilers/polyblocks.yaml index 2744cef..e795c6b 100644 --- a/compilers/polyblocks.yaml +++ b/compilers/polyblocks.yaml @@ -8,6 +8,7 @@ polyblocks: - mlir - proprietary + temperature: hot url: https://www.polymagelabs.com/technology/ description: | diff --git a/compilers/shark.yaml b/compilers/shark.yaml index f13fac0..f50cbc2 100644 --- a/compilers/shark.yaml +++ b/compilers/shark.yaml @@ -9,6 +9,7 @@ shark: - amd - apache-2.0 + temperature: hot url: https://github.com/nod-ai/SHARK/tree/main description: | diff --git a/compilers/stable-fast.yaml b/compilers/stable-fast.yaml new file mode 100644 index 0000000..ba6202b --- /dev/null +++ b/compilers/stable-fast.yaml @@ -0,0 +1,43 @@ +stable-fast: + name: "stable-fast" + + image_url: https://docs.oneflow.org/en/master/assets/product-layer.png + + tags: + - compilers + - framework + - apache-2.0 + + temperature: hot + url: https://github.com/chengzeyi/stable-fast + + description: | + StableFast is a cutting-edge, ultra-lightweight inference optimization framework designed + specifically for HuggingFace Diffusers on NVIDIA GPUs. It stands out for its exceptional + state-of-the-art (SOTA) inference performance on a wide range of diffuser models, including + the latest StableVideoDiffusionPipeline. One of its most notable features is its rapid model + compilation capability, which significantly outpaces other frameworks like TensorRT or AITemplate + by reducing compilation time from minutes to mere seconds. StableFast supports dynamic shapes, LoRA + (Low-Rank Adaptation), and ControlNet, offering a broad range of functionalities. It incorporates + advanced techniques such as CUDNN Convolution Fusion, low precision and fused GEMM, fused Linear GEGLU, + optimized NHWC & fused GroupNorm, and CUDA Graph and Fused Multihead Attention optimizations. This makes + it a highly versatile and efficient tool for developers. The framework is compatible with various versions + of HuggingFace Diffusers and PyTorch, ensuring broad applicability. Currently tested on Linux and WSL2 in + Windows, StableFast requires PyTorch with CUDA support and specific versions of related tools like xformers + and triton. The ongoing development of StableFast focuses on maintaining its position as a leading inference + optimization framework, with an emphasis on enhancing speed and reducing VRAM usage for transformers. This + commitment to continuous improvement underlines its utility in the rapidly evolving field of deep learning + optimization. + + features: + - "Rapid Model Compilation" + - "Supports Dynamic Shape" + - "Compatible with LoRA and ControlNet" + - "CUDNN Convolution Fusion" + - "Low Precision & Fused GEMM Operations" + - "Fused Linear GEGLU" + - "Optimized NHWC & Fused GroupNorm" + - "Enhanced TorchScript Tracing" + - "CUDA Graph Support" + - "Fused Multihead Attention" + - "Broad Compatibility with PyTorch and HuggingFace Diffusers" diff --git a/compilers/taco.yaml b/compilers/taco.yaml index 5490465..fadf671 100644 --- a/compilers/taco.yaml +++ b/compilers/taco.yaml @@ -7,6 +7,7 @@ taco: - compilers - mit + temperature: cold url: http://tensor-compiler.org description: | diff --git a/compilers/tensor-comprehensions.yaml b/compilers/tensor-comprehensions.yaml index 2683ba3..4451ba7 100644 --- a/compilers/tensor-comprehensions.yaml +++ b/compilers/tensor-comprehensions.yaml @@ -7,6 +7,7 @@ tiramisu: - compilers - apache-2.0 + temperature: cold url: https://github.com/facebookresearch/TensorComprehensions description: | diff --git a/compilers/tensorrt-llm.yaml b/compilers/tensorrt-llm.yaml index 7b5ade2..ebb0db9 100644 --- a/compilers/tensorrt-llm.yaml +++ b/compilers/tensorrt-llm.yaml @@ -10,6 +10,7 @@ tensorrt-llm: - inference-optimizer - apache-2.0 + temperature: hot url: https://github.com/NVIDIA/TensorRT-LLM description: | diff --git a/compilers/tensorrt.yaml b/compilers/tensorrt.yaml index 9fd78a0..ca48af4 100644 --- a/compilers/tensorrt.yaml +++ b/compilers/tensorrt.yaml @@ -9,6 +9,7 @@ tensorrt: - inference-optimizer - apache-2.0 + temperature: hot url: https://developer.nvidia.com/tensorrt description: | diff --git a/compilers/tinygrad.yaml b/compilers/tinygrad.yaml index 6a0ad44..b69e61e 100644 --- a/compilers/tinygrad.yaml +++ b/compilers/tinygrad.yaml @@ -8,6 +8,7 @@ tinygrad: - framework - mit + temperature: hot url: https://tinygrad.org description: | diff --git a/compilers/tiramisu.yaml b/compilers/tiramisu.yaml index 349a5b6..1ea039d 100644 --- a/compilers/tiramisu.yaml +++ b/compilers/tiramisu.yaml @@ -7,6 +7,7 @@ tiramisu: - compilers - mit + temperature: cold url: https://tiramisu-compiler.org description: | diff --git a/compilers/torch-mlir.yaml b/compilers/torch-mlir.yaml index fe84d03..e1cb441 100644 --- a/compilers/torch-mlir.yaml +++ b/compilers/torch-mlir.yaml @@ -9,6 +9,7 @@ torch-mlir: - other-license - bsd-3-clause + temperature: hot url: https://github.com/llvm/torch-mlir description: | diff --git a/compilers/triton.yaml b/compilers/triton.yaml index 5e75a33..d00c482 100644 --- a/compilers/triton.yaml +++ b/compilers/triton.yaml @@ -9,6 +9,7 @@ triton: - nvidia - mit + temperature: hot url: https://openai.com/research/triton description: | diff --git a/compilers/tvm.yaml b/compilers/tvm.yaml index 02151ad..2bec7cc 100644 --- a/compilers/tvm.yaml +++ b/compilers/tvm.yaml @@ -7,6 +7,7 @@ tvm: - compilers - apache-2.0 + temperature: hot url: https://tvm.apache.org description: | diff --git a/compilers/vllm.yaml b/compilers/vllm.yaml index 0a185f0..150c1de 100644 --- a/compilers/vllm.yaml +++ b/compilers/vllm.yaml @@ -10,6 +10,7 @@ vllm: - high-throughput - apache-2.0 + temperature: hot url: https://vllm.ai description: | diff --git a/compilers/xla.yaml b/compilers/xla.yaml index 3a8c68a..bc9c17a 100644 --- a/compilers/xla.yaml +++ b/compilers/xla.yaml @@ -8,6 +8,7 @@ xla: - mlir - apache-2.0 + temperature: hot url: https://www.tensorflow.org/xla description: |