Rate this Page

★ ★ ★ ★ ★

recipes/torch_export_aoti_python

Run in Google Colab

Note

Go to the end to download the full example code.

`torch.export` AOTInductor Tutorial for Python runtime (Beta)#

Created On: Aug 23, 2024 | Last Updated: Jan 24, 2025 | Last Verified: Nov 05, 2024

Author: Ankith Gunapal, Bin Bao, Angela Yi

Warning

torch._inductor.aoti_compile_and_package and torch._inductor.aoti_load_package are in Beta status and are subject to backwards compatibility breaking changes. This tutorial provides an example of how to use these APIs for model deployment using Python runtime.

It has been shown previously how AOTInductor can be used to do Ahead-of-Time compilation of PyTorch exported models by creating an artifact that can be run in a non-Python environment. In this tutorial, you will learn an end-to-end example of how to use AOTInductor for Python runtime.

Contents

Prerequisites #

PyTorch 2.6 or later
Basic understanding of torch.export and AOTInductor
Complete the AOTInductor: Ahead-Of-Time Compilation for Torch.Export-ed Models tutorial

What you will learn #

How to use AOTInductor for Python runtime.
How to use torch._inductor.aoti_compile_and_package() along with torch.export.export() to generate a compiled artifact
How to load and run the artifact in a Python runtime using torch._export.aot_load().
When to you use AOTInductor with a Python runtime

Model Compilation #

We will use the TorchVision pretrained ResNet18 model as an example.

The first step is to export the model to a graph representation using torch.export.export(). To learn more about using this function, you can check out the docs or the tutorial.

Once we have exported the PyTorch model and obtained an ExportedProgram, we can apply torch._inductor.aoti_compile_and_package() to AOTInductor to compile the program to a specified device, and save the generated contents into a “.pt2” artifact.

Note

This API supports the same available options that torch.compile() has, such as mode and max_autotune (for those who want to enable CUDA graphs and leverage Triton based matrix multiplications and convolutions)

import os
import torch
import torch._inductor
from torchvision.models import ResNet18_Weights, resnet18

model = resnet18(weights=ResNet18_Weights.DEFAULT)
model.eval()

with torch.inference_mode():
    inductor_configs = {}

    if torch.cuda.is_available():
        device = "cuda"
        inductor_configs["max_autotune"] = True
    else:
        device = "cpu"

    model = model.to(device=device)
    example_inputs = (torch.randn(2, 3, 224, 224, device=device),)

    exported_program = torch.export.export(
        model,
        example_inputs,
    )
    path = torch._inductor.aoti_compile_and_package(
        exported_program,
        package_path=os.path.join(os.getcwd(), "resnet18.pt2"),
        inductor_configs=inductor_configs
    )

Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /var/lib/ci-user/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth

  0%|          | 0.00/44.7M [00:00<?, ?B/s]
 93%|█████████▎| 41.5M/44.7M [00:00<00:00, 435MB/s]
100%|██████████| 44.7M/44.7M [00:00<00:00, 435MB/s]
/usr/lib/python3.10/copyreg.py:101: FutureWarning:

`isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead.

/usr/local/lib/python3.10/dist-packages/torch/_inductor/compile_fx.py:321: UserWarning:

TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.

/usr/local/lib/python3.10/dist-packages/torch/_inductor/select_algorithm.py:3464: UserWarning:

TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()

Autotune Choices Stats:
{"num_choices": 7, "num_triton_choices": 6, "best_kernel": "convolution", "best_time": 0.0870399996638298, "best_triton_pos": 1, "best_triton_time": 0.11878400295972824, "best_triton_kernel": "triton_convolution2d_0", "best_triton_kernel_desc": "ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=7, KERNEL_W=7, PADDING_H=3, PADDING_W=3, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4"}
AUTOTUNE convolution(2x3x224x224, 64x3x7x7)
strides: [150528, 1, 672, 3], [147, 1, 21, 3]
dtypes: torch.float32, torch.float32
  convolution 0.0870 ms 100.0%
  triton_convolution2d_0 0.1188 ms 73.3% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=7, KERNEL_W=7, PADDING_H=3, PADDING_W=3, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_4 0.1536 ms 56.7% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=7, KERNEL_W=7, PADDING_H=3, PADDING_W=3, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_3 0.1802 ms 48.3% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=128, BLOCK_N=64, GROUPS=1, KERNEL_H=7, KERNEL_W=7, PADDING_H=3, PADDING_W=3, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_1 0.2047 ms 42.5% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=7, KERNEL_W=7, PADDING_H=3, PADDING_W=3, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_5 0.2202 ms 39.5% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=7, KERNEL_W=7, PADDING_H=3, PADDING_W=3, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_2 0.3011 ms 28.9% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=1024, BLOCK_N=16, GROUPS=1, KERNEL_H=7, KERNEL_W=7, PADDING_H=3, PADDING_W=3, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=1, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 0.2795 seconds and 0.0017 seconds precompiling for 7 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "triton_convolution2d_11", "best_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8", "best_time": 0.03686400130391121, "best_triton_pos": 0}
AUTOTUNE convolution(2x64x56x56, 64x64x3x3)
strides: [200704, 1, 3584, 64], [576, 1, 192, 64]
dtypes: torch.float32, torch.float32
  triton_convolution2d_11 0.0369 ms 100.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_6 0.0430 ms 85.7% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_9 0.0430 ms 85.7% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_10 0.0430 ms 85.7% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_7 0.0594 ms 62.1% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_12 0.0604 ms 61.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  convolution 0.0788 ms 46.8%
  triton_convolution2d_8 0.1178 ms 31.3% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=1024, BLOCK_N=16, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=1, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 0.0974 seconds and 0.0006 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "triton_convolution2d_38", "best_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4", "best_time": 0.03788800165057182, "best_triton_pos": 0}
AUTOTUNE convolution(2x64x56x56, 128x64x3x3)
strides: [200704, 1, 3584, 64], [576, 1, 192, 64]
dtypes: torch.float32, torch.float32
  triton_convolution2d_38 0.0379 ms 100.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_39 0.0430 ms 88.1% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_34 0.0471 ms 80.4% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_40 0.0604 ms 62.7% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_35 0.0614 ms 61.7% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_37 0.0655 ms 57.8% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  convolution 0.0707 ms 53.6%
  triton_convolution2d_36 0.1147 ms 33.0% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=1024, BLOCK_N=16, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=1, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 0.1235 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "triton_convolution2d_45", "best_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4", "best_time": 0.048128001391887665, "best_triton_pos": 0}
AUTOTUNE convolution(2x128x28x28, 128x128x3x3)
strides: [100352, 1, 3584, 128], [1152, 1, 384, 128]
dtypes: torch.float32, torch.float32
  triton_convolution2d_45 0.0481 ms 100.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_46 0.0758 ms 63.5% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  convolution 0.0788 ms 61.0%
  triton_convolution2d_41 0.0881 ms 54.7% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_42 0.1075 ms 44.8% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_47 0.1126 ms 42.7% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_44 0.1208 ms 39.8% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_43 0.2222 ms 21.7% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=1024, BLOCK_N=16, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=1, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 0.1581 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "triton_convolution2d_49", "best_kernel_desc": "ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4", "best_time": 0.018432000651955605, "best_triton_pos": 0}
AUTOTUNE convolution(2x64x56x56, 128x64x1x1)
strides: [200704, 1, 3584, 64], [64, 1, 1, 1]
dtypes: torch.float32, torch.float32
  triton_convolution2d_49 0.0184 ms 100.0% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_52 0.0256 ms 72.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_48 0.0266 ms 69.2% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=128, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_51 0.0266 ms 69.2% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_53 0.0266 ms 69.2% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=128, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_54 0.0266 ms 69.2% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_50 0.0276 ms 66.7% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=1024, BLOCK_N=16, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=1, num_warps=8
  convolution 0.0584 ms 31.6%
SingleProcess AUTOTUNE benchmarking takes 0.1088 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "triton_convolution2d_73", "best_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4", "best_time": 0.048128001391887665, "best_triton_pos": 0}
AUTOTUNE convolution(2x128x28x28, 256x128x3x3)
strides: [100352, 1, 3584, 128], [1152, 1, 384, 128]
dtypes: torch.float32, torch.float32
  triton_convolution2d_73 0.0481 ms 100.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  convolution 0.0809 ms 59.5%
  triton_convolution2d_75 0.1157 ms 41.6% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_70 0.1188 ms 40.5% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_72 0.1239 ms 38.8% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_74 0.1290 ms 37.3% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_69 0.1331 ms 36.2% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_71 0.1792 ms 26.9% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=1024, BLOCK_N=16, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=1, num_warps=8
SingleProcess AUTOTUNE benchmarking takes 0.1734 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "convolution", "best_time": 0.08499199897050858, "best_triton_pos": 1, "best_triton_time": 0.09011200070381165, "best_triton_kernel": "triton_convolution2d_80", "best_triton_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4"}
AUTOTUNE convolution(2x256x14x14, 256x256x3x3)
strides: [50176, 1, 3584, 256], [2304, 1, 768, 256]
dtypes: torch.float32, torch.float32
  convolution 0.0850 ms 100.0%
  triton_convolution2d_80 0.0901 ms 94.3% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_77 0.2089 ms 40.7% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_78 0.2150 ms 39.5% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=512, BLOCK_N=16, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=1, num_warps=8
  triton_convolution2d_82 0.2202 ms 38.6% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_79 0.2324 ms 36.6% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_81 0.2519 ms 33.7% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_76 0.2621 ms 32.4% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 0.2315 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "triton_convolution2d_84", "best_kernel_desc": "ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4", "best_time": 0.01740800030529499, "best_triton_pos": 0}
AUTOTUNE convolution(2x128x28x28, 256x128x1x1)
strides: [100352, 1, 3584, 128], [128, 1, 1, 1]
dtypes: torch.float32, torch.float32
  triton_convolution2d_84 0.0174 ms 100.0% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_87 0.0256 ms 68.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_83 0.0276 ms 63.0% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_86 0.0276 ms 63.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_88 0.0287 ms 60.7% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_89 0.0297 ms 58.6% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_85 0.0348 ms 50.0% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=1024, BLOCK_N=16, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=1, num_warps=8
  convolution 0.0563 ms 30.9%
SingleProcess AUTOTUNE benchmarking takes 0.1066 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "convolution", "best_time": 0.08908800035715103, "best_triton_pos": 1, "best_triton_time": 0.09318400174379349, "best_triton_kernel": "triton_convolution2d_108", "best_triton_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4"}
AUTOTUNE convolution(2x256x14x14, 512x256x3x3)
strides: [50176, 1, 3584, 256], [2304, 1, 768, 256]
dtypes: torch.float32, torch.float32
  convolution 0.0891 ms 100.0%
  triton_convolution2d_108 0.0932 ms 95.6% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_106 0.2109 ms 42.2% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=512, BLOCK_N=16, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=1, num_warps=8
  triton_convolution2d_110 0.2232 ms 39.9% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_105 0.2355 ms 37.8% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_107 0.2396 ms 37.2% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_109 0.2529 ms 35.2% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_104 0.2601 ms 34.3% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=2, STRIDE_W=2, UNROLL=False, num_stages=2, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 0.2323 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "convolution", "best_time": 0.09932799637317657, "best_triton_pos": 1, "best_triton_time": 0.17612800002098083, "best_triton_kernel": "triton_convolution2d_115", "best_triton_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4"}
AUTOTUNE convolution(2x512x7x7, 512x512x3x3)
strides: [25088, 1, 3584, 512], [4608, 1, 1536, 512]
dtypes: torch.float32, torch.float32
  convolution 0.0993 ms 100.0%
  triton_convolution2d_115 0.1761 ms 56.4% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_113 0.2150 ms 46.2% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=128, BLOCK_N=16, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=1, num_warps=8
  triton_convolution2d_117 0.2796 ms 35.5% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_112 0.3492 ms 28.4% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=128, BLOCK_N=64, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
  triton_convolution2d_114 0.4608 ms 21.6% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_116 0.4925 ms 20.2% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=8
  triton_convolution2d_111 0.5140 ms 19.3% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=3, KERNEL_W=3, PADDING_H=1, PADDING_W=1, STRIDE_H=1, STRIDE_W=1, UNROLL=False, num_stages=2, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 0.2462 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 8, "num_triton_choices": 7, "best_kernel": "triton_convolution2d_122", "best_kernel_desc": "ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4", "best_time": 0.026655999943614006, "best_triton_pos": 0}
AUTOTUNE convolution(2x256x14x14, 512x256x1x1)
strides: [50176, 1, 3584, 256], [256, 1, 1, 1]
dtypes: torch.float32, torch.float32
  triton_convolution2d_122 0.0267 ms 100.0% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_118 0.0298 ms 89.6% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_124 0.0307 ms 86.8% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_119 0.0358 ms 74.4% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=256, BLOCK_N=64, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=4
  triton_convolution2d_120 0.0358 ms 74.4% ALLOW_TF32=True, BLOCK_K=16, BLOCK_M=512, BLOCK_N=16, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=1, num_warps=8
  triton_convolution2d_121 0.0399 ms 66.7% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=128, BLOCK_N=128, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  triton_convolution2d_123 0.0420 ms 63.5% ALLOW_TF32=True, BLOCK_K=32, BLOCK_M=64, BLOCK_N=256, GROUPS=1, KERNEL_H=1, KERNEL_W=1, PADDING_H=0, PADDING_W=0, STRIDE_H=2, STRIDE_W=2, UNROLL=True, num_stages=2, num_warps=8
  convolution 0.0553 ms 48.2%
SingleProcess AUTOTUNE benchmarking takes 0.1074 seconds and 0.0002 seconds precompiling for 8 choices
Autotune Choices Stats:
{"num_choices": 13, "num_triton_choices": 12, "best_kernel": "triton_mm_142", "best_kernel_desc": "ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=16, BLOCK_N=32, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=5, num_warps=2", "best_time": 0.027648000046610832, "best_triton_pos": 0}
AUTOTUNE addmm(2x1000, 2x512, 512x1000)
strides: [0, 1], [512, 1], [1, 512]
dtypes: torch.float32, torch.float32, torch.float32
  triton_mm_142 0.0276 ms 100.0% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=16, BLOCK_N=32, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=5, num_warps=2
  triton_mm_140 0.0317 ms 87.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=128, BLOCK_M=16, BLOCK_N=32, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=2, num_warps=2
  triton_mm_145 0.0389 ms 71.1% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=64, BLOCK_M=16, BLOCK_N=64, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=3, num_warps=4
  triton_mm_141 0.0410 ms 67.5% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=16, BLOCK_N=64, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=5, num_warps=4
  addmm 0.0420 ms 65.9%
  triton_mm_139 0.0420 ms 65.9% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=16, BLOCK_M=16, BLOCK_N=32, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=1, num_warps=2
  triton_mm_148 0.0430 ms 64.3% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=16, BLOCK_N=64, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=3, num_warps=4
  triton_mm_149 0.0440 ms 62.8% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=16, BLOCK_N=64, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=4, num_warps=4
  triton_mm_143 0.0451 ms 61.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=16, BLOCK_M=16, BLOCK_N=64, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=2, num_warps=4
  triton_mm_144 0.0451 ms 61.4% ACC_TYPE='tl.float32', ALLOW_TF32=False, BLOCK_K=32, BLOCK_M=16, BLOCK_N=64, EVEN_K=True, GROUP_M=8, USE_FAST_ACCUM=False, num_stages=2, num_warps=4
SingleProcess AUTOTUNE benchmarking takes 0.3733 seconds and 0.0002 seconds precompiling for 13 choices

The result of aoti_compile_and_package() is an artifact “resnet18.pt2” which can be loaded and executed in Python and C++.

The artifact itself contains a bunch of AOTInductor generated code, such as a generated C++ runner file, a shared library compiled from the C++ file, and CUDA binary files, aka cubin files, if optimizing for CUDA.

Structure-wise, the artifact is a structured .zip file, with the following specification:

We can use the following command to inspect the artifact contents:

$ unzip -l resnet18.pt2

Archive:  resnet18.pt2
  Length      Date    Time    Name
---------  ---------- -----   ----
        1  01-08-2025 16:40   version
        3  01-08-2025 16:40   archive_format
    10088  01-08-2025 16:40   data/aotinductor/model/cagzt6akdaczvxwtbvqe34otfe5jlorktbqlojbzqjqvbfsjlge4.cubin
    17160  01-08-2025 16:40   data/aotinductor/model/c6oytfjmt5w4c7onvtm6fray7clirxt7q5xjbwx3hdydclmwoujz.cubin
    16616  01-08-2025 16:40   data/aotinductor/model/c7ydp7nocyz323hij4tmlf2kcedmwlyg6r57gaqzcsy3huneamu6.cubin
    17776  01-08-2025 16:40   data/aotinductor/model/cyqdf46ordevqhiddvpdpp3uzwatfbzdpl3auj2nx23uxvplnne2.cubin
    10856  01-08-2025 16:40   data/aotinductor/model/cpzfebfgrusqslui7fxsuoo4tvwulmrxirc5tmrpa4mvrbdno7kn.cubin
    14608  01-08-2025 16:40   data/aotinductor/model/c5ukeoz5wmaszd7vczdz2qhtt6n7tdbl3b6wuy4rb2se24fjwfoy.cubin
    11376  01-08-2025 16:40   data/aotinductor/model/csu3nstcp56tsjfycygaqsewpu64l5s6zavvz7537cm4s4cv2k3r.cubin
    10984  01-08-2025 16:40   data/aotinductor/model/cp76lez4glmgq7gedf2u25zvvv6rksv5lav4q22dibd2zicbgwj3.cubin
    14736  01-08-2025 16:40   data/aotinductor/model/c2bb5p6tnwz4elgujqelsrp3unvkgsyiv7xqxmpvuxcm4jfl7pc2.cubin
    11376  01-08-2025 16:40   data/aotinductor/model/c6eopmb2b4ngodwsayae4r5q6ni3jlfogfbdk3ypg56tgpzhubfy.cubin
    11624  01-08-2025 16:40   data/aotinductor/model/chmwe6lvoekzfowdbiizitm3haiiuad5kdm6sd2m6mv6dkn2zk32.cubin
    15632  01-08-2025 16:40   data/aotinductor/model/c3jop5g344hj3ztsu4qm6ibxyaaerlhkzh2e6emak23rxfje6jam.cubin
    25472  01-08-2025 16:40   data/aotinductor/model/chaiixybeiuuitm2nmqnxzijzwgnn2n7uuss4qmsupgblfh3h5hk.cubin
   139389  01-08-2025 16:40   data/aotinductor/model/cvk6qzuybruhwxtfblzxiov3rlrziv5fkqc4mdhbmantfu3lmd6t.cpp
       27  01-08-2025 16:40   data/aotinductor/model/cvk6qzuybruhwxtfblzxiov3rlrziv5fkqc4mdhbmantfu3lmd6t_metadata.json
 47195424  01-08-2025 16:40   data/aotinductor/model/cvk6qzuybruhwxtfblzxiov3rlrziv5fkqc4mdhbmantfu3lmd6t.so
---------                     -------
 47523148                     18 files

Model Inference in Python #

To load and run the artifact in Python, we can use torch._inductor.aoti_load_package().

import os
import torch
import torch._inductor

model_path = os.path.join(os.getcwd(), "resnet18.pt2")

compiled_model = torch._inductor.aoti_load_package(model_path)
example_inputs = (torch.randn(2, 3, 224, 224, device=device),)

with torch.inference_mode():
    output = compiled_model(example_inputs)

When to use AOTInductor with a Python Runtime #

There are mainly two reasons why one would use AOTInductor with a Python Runtime:

torch._inductor.aoti_compile_and_package generates a singular serialized artifact. This is useful for model versioning for deployments and tracking model performance over time.
With torch.compile() being a JIT compiler, there is a warmup cost associated with the first compilation. Your deployment needs to account for the compilation time taken for the first inference. With AOTInductor, the compilation is done ahead of time using torch.export.export and torch._inductor.aoti_compile_and_package. At deployment time, after loading the model, running inference does not have any additional cost.

The section below shows the speedup achieved with AOTInductor for first inference

We define a utility function timed to measure the time taken for inference

import time
def timed(fn):
    # Returns the result of running `fn()` and the time it took for `fn()` to run,
    # in seconds. We use CUDA events and synchronization for accurate
    # measurement on CUDA enabled devices.
    if torch.cuda.is_available():
        start = torch.cuda.Event(enable_timing=True)
        end = torch.cuda.Event(enable_timing=True)
        start.record()
    else:
        start = time.time()

    result = fn()
    if torch.cuda.is_available():
        end.record()
        torch.cuda.synchronize()
    else:
        end = time.time()

    # Measure time taken to execute the function in miliseconds
    if torch.cuda.is_available():
        duration = start.elapsed_time(end)
    else:
        duration = (end - start) * 1000

    return result, duration

Lets measure the time for first inference using AOTInductor

torch._dynamo.reset()

model = torch._inductor.aoti_load_package(model_path)
example_inputs = (torch.randn(1, 3, 224, 224, device=device),)

with torch.inference_mode():
    _, time_taken = timed(lambda: model(example_inputs))
    print(f"Time taken for first inference for AOTInductor is {time_taken:.2f} ms")

Time taken for first inference for AOTInductor is 3.30 ms

Lets measure the time for first inference using torch.compile

torch._dynamo.reset()

model = resnet18(weights=ResNet18_Weights.DEFAULT).to(device)
model.eval()

model = torch.compile(model)
example_inputs = torch.randn(1, 3, 224, 224, device=device)

with torch.inference_mode():
    _, time_taken = timed(lambda: model(example_inputs))
    print(f"Time taken for first inference for torch.compile is {time_taken:.2f} ms")

Time taken for first inference for torch.compile is 4133.58 ms

We see that there is a drastic speedup in first inference time using AOTInductor compared to torch.compile

Conclusion #

In this recipe, we have learned how to effectively use the AOTInductor for Python runtime by compiling and loading a pretrained ResNet18 model. This process demonstrates the practical application of generating a compiled artifact and running it within a Python environment. We also looked at the advantage of using AOTInductor in model deployments, with regards to speed up in first inference time.

Total running time of the script: (0 minutes 28.851 seconds)

`torch.export` AOTInductor Tutorial for Python runtime (Beta)#

Prerequisites #

What you will learn #

Model Compilation #

Model Inference in Python #

When to use AOTInductor with a Python Runtime #

Conclusion #

Docs

Tutorials

Resources

torch.export AOTInductor Tutorial for Python runtime (Beta)#

Prerequisites#

What you will learn#

Model Compilation#

Model Inference in Python#

When to use AOTInductor with a Python Runtime#

Conclusion#

Docs

Tutorials

Resources

`torch.export` AOTInductor Tutorial for Python runtime (Beta)#

Prerequisites #

What you will learn #

Model Compilation #

Model Inference in Python #

When to use AOTInductor with a Python Runtime #

Conclusion #