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.

facebookincubator/AITemplate Issues

model_interface.cu:231: Error: Constant pretrained_model_patch_embed_proj_weight was not set! Set the value with set_constant.
Closed 2 months ago1
Got cutlass error: Error Internal at: 214
Updated 2 months ago1
Can i remove vae?
Closed 2 months ago1
Docker image fails to build due to python dependency issue
Updated 2 months ago
error during inferencing: Error: Constant embeddings_token_embedding_weight was not set! Set the value with set_constant.
Updated 3 months ago2
Does Concatenate order matters?
Updated 3 months ago1
windows platform cannot link _binary_constants_bin_end and _binary_constants_bin_start
Closed 3 months ago5
Model is successfully compiled, but OOM when loading
Updated 4 months ago
multi-gpu at runtime error
Updated 4 months ago4
Does AIT handle if/else in forward function?
Updated 4 months ago1
<class 'src.pipeline_stable_diffusion_ait.StableDiffusionAITPipeline'> is incorrectly implemented. Expected {'feature_extractor', 'scheduler', 'tokenizer', 'text_encoder', 'safety_checker', 'unet', 'vae'} to be defined
Updated 5 months ago3
`Unsupported workload for this conv2d specialization` when using dynamic shape together with permute
Closed 5 months ago6
gcc: internal compiler error
Closed 5 months ago2
`Fatal Python error: Floating point exception` with `run_with_tensors`
Closed 6 months ago3
Failed to compile the controlnet: thepowefuldeez/sd21-controlnet-canny
Updated 6 months ago1
gemm_gen_profiler() takes 3 positional arguments but 4 were given
Closed 6 months ago2
Low performance from unnecessary permutations
Updated 6 months ago18
Stable Diffusion (GLIGEN) Download Error
Updated 6 months ago4
Building rocm docker image is very slow. Is there a prebuilt docker image?
Updated 6 months ago1
Stable Diffusion SD XL support
Updated 7 months ago10
complie controlnet error
Closed 7 months ago2
Compile Diffusers Community Pipelines
Updated 7 months ago1
Would it be possible to port RotaryEmbedding to AITemplate
Closed 8 months ago6
AIT seems not able to achieve max along a given dim
Closed 8 months ago9
fx2ait install error -- c10::StorageImpl has no member named mutable_data
Closed 8 months ago1
AMD MI210 ResNet test error
Updated 8 months ago3
VAEdecode and AITemplate dont get along, slower latent decoding
Updated 8 months ago3
GFX1100 Support
Updated 8 months ago3
sm90/sm90a Hopper architecture Incompatibility
Closed 8 months ago3
I try to make it work on gfx1030,but
Closed 8 months ago1
Problem running ControlNet example with AITemplate
Closed 9 months ago1
op b2b_bmm report an error "no module named cutlass_lib'
Updated 9 months ago3
fx2ait low performance
Updated 9 months ago1
Ops unit tests fail on ROCm
Updated 9 months ago1
Does AIT have FlashAttention support for AMD ROCm?
Closed 9 months ago1
making template of sdxl
Closed 9 months ago1
Add Support for intel arc GPUs 🥹
Closed 9 months ago1
Does AIT support BF16 inference now?
Updated 9 months ago4
Could AITemplate support mix precision inference?
Updated 9 months ago2
Compile error while transplanting flash attention2 to AITemplate
Closed 10 months ago3
Slicing tensor along a fixed dimension
Closed 10 months ago2
Any plan on supporting Flash Attention 2
Updated 10 months ago1
Demo for using aitemplate for transformers based LLMs
Updated 10 months ago
Error in installing fx2ait setup.py
Updated 10 months ago
Option for choosing fp32 gemm backend implementation
Closed 10 months ago4
Dynamic Resolution for Control-Net
Closed 10 months ago1
How do I make stable diffusion with AIT work with any latent resolution?
Updated 10 months ago5
Error in compiling dynamic resolution on SD 1.5
Closed 10 months ago2
fx2ait has torch version limitations?
Closed 10 months ago3
The input/output of each kernel in stable diffustion
Updated 10 months ago1