facebookincubator/AITemplate
AITemplate is a Python framework which renders neural network into high performance CUDA/HIP C++ code. Specialized for FP16 TensorCore (NVIDIA GPU) and MatrixCore (AMD GPU) inference.
PythonApache-2.0
Issues
- 1
- 10
Stable Diffusion SD XL support
#851 opened - 3
fx2ait has torch version limitations?
#850 opened - 1
- 1
- 0
Disscusion about improving compile time.
#828 opened - 2
Plans for CPU backend
#827 opened - 7
Does Conv2d Kernels Support Float32 on SM75?
#823 opened - 2
Creating a conda/pip package
#821 opened - 3
Question: Same Operator in Different Modules
#818 opened - 2
- 2
Error for compiling controlnet
#808 opened - 1
- 0
- 9
How to compile instruct pix2pix using AIT
#802 opened - 12
Error when run stable diffusion with controlnet.
#797 opened - 0
- 5
The size of tensor a (64) must match the size of tensor b (128) at non-singleton dimension 3
#789 opened - 4
- 3
Error compiling stable diffusion examples
#780 opened - 2
- 3
Cannot Build fx2ait with setup.py
#778 opened - 4
- 2
[fx2ait] "Input shapes of the elementwise op are not compatible" while running bert model
#753 opened - 1
llama support
#752 opened - 7
Stable Diffusion 2.1 768x786 demo.py failed
#751 opened - 3
- 4
Error when running compile_alt.py in stable diffusion example: list index out of range in conv2d
#742 opened - 3
compilation fails on example 05
#738 opened - 3
- 5
Hidden embedding size in compile_controlnet
#730 opened - 7
Stable Diffusion demo fails to complile VAE
#723 opened - 10
- 2
Compilation for multiple GPUs
#719 opened - 0
Attention mask in Bert
#712 opened - 1
- 4
StableDiffusion benchmark script cannot handle larger batch sizes even when compiling at size
#685 opened - 2
[Feature Request] Example for MIDAS
#678 opened - 1
Combine stable diffusion pipelines
#659 opened - 1
stable diffusion img2img fails
#652 opened - 9
Stable Diffusion example is failing
#650 opened - 2
Slow nn.Linear on MI250
#648 opened - 2
- 1
d2 这样成功了吗？
#618 opened - 3
Tensor Parallelism
#616 opened - 1
python scripts/compile.py error
#613 opened - 2
- 1
- 0
How to use the exported library with c++?
#590 opened - 5