/BladeDISC

BladeDISC is an end-to-end DynamIc Shape Compiler project for machine learning workloads.

Primary LanguageC++Apache License 2.0Apache-2.0

BladeDISC Introduction

We're hiring!🔥🔥🔥

We're always looking for candicates to join dev team. Your're the one we're searching for long:

  • 🥷 If you are an compiler or AI enthusiasts.
  • ⭐️ or if you are experienced in optimization on CPUs and GPUs.
  • ⚙️ or if you wanna build an unified and automated compiler to optimize both inference and training workload.
  • 🤿 or if you are using BladeDISC in production or research projects, and wanna have a deeper dive into it.
  • ✄ or you wanna build cutting-edge infrastructure in the AIGC era.

Please contact us via email or Dingtalk at the bottom of page.⬇️⬇️⬇️

What's New

Overview

BladeDISC is an end-to-end DynamIc Shape Compiler project for machine learning workloads, which is one of the key components of Alibaba's PAI-Blade. BladeDISC provides general, transparent, and ease-of-use performance optimization for TensorFlow/PyTorch workloads on GPGPU and CPU backends. The architecture natively supports dynamic shape workloads, with many considerations in the performance of both static and dynamic shape scenarios. It also supports multiple and flexible deployment solutions, including both Plugin Mode inside TensorFlow/PyTorch runtime, and Standalone Mode for AOT standalone execution. The project is based on MLIR and highly related to mlir-hlo project.

Refer to our website for more information, including the setup tutorial, developer guide, demo examples and documents for developers.

Features and Roadmap

Frontend Framework Support Matrix

TensorFlow [1] PyTorch [2]
Inference Yes Yes
Training Yes [3] Ongoing

[1] TensorFlow 1.12, 1.15, 2.4 & 2.5 are supported and fully verified. For other versions, some slight work on adaptation might be needed.

[2] PyTorch version >= 1.6.0 has been fully verified.

[3] Although supported, there's much room for improvement on Op coverage for training workloads.

Backend Support Matrix

Status
Nvidia GPU Yes [1]
AMD GPU Yes
Hygon DCU Yes
X86 Yes
AArch64 Yes

[1] Support for CUDA below 11.0 has been deprecated officially since Aug 2022.

Deployment Solutions

  • Plugin Mode - BladeDISC works as a plugin of TensorFlow or PyTorch. Only the supported Ops are clustered and compiled, and the unsupported ones will be executed by the original TensorFlow or PyTorch runtime. We recommend this mode to most of the users for its transparency and ease of use.

  • Standalone Mode - In Standalone mode, the input workload will be compiled into a binary that can be executed by itself, aka, does not rely on a TensorFlow or PyTorch runtime. In this mode, all ops must be supported.

Numbers of Typical Workloads

By evaluating BladeDISC using a set of typical machine learning workloads for production purposes, BladeDISC shows up to 6.95x speedup compared with PyTorch. Moreover, compared to static optimizing compilers (i.e., XLA and TensorRT), BladeDISC shows comparable or even better performance.

Fig.1 End-to-end Performance of BladeDISC and baselines. Note that some baselines fail to optimize ViT model.

Advantage in Dynamic Shape Workloads

Specifically, for the BERT large inference on T4 GPU, we provide in the examples, static compiler optimization (XLA) shows severe performance degradation due to its compilation overhead, while BladeDISC shows a 1.75x speedup.

TensorFlow XLA BladeDISC
1.78 s 41.69s 1.02s
1X 1.75X

API QuickView

For TensorFlow Users

Only two lines of code are needed on native TensorFlow program as the following:

import numpy as np
import tensorflow as tf

## enable BladeDISC on TensorFlow program
import blade_disc_tf as disc
disc.enable()

## construct TensorFlow Graph and run it
g = tf.Graph()
with g.as_default():
    ...
    with tf.session as sess:
        sess.run(...)

For more information, please refer to QuickStart for TensorFlow Users

For PyTorch Users

PyTorch users only need the following few lines of code to enable BladeDISC:

import torch_blade
# construct PyTorch Module
class MyModule(nn.Module):
    ...

module = MyModule().eval()

with torch.no_grad():
    # blade_module is the optimized module by BladeDISC
    blade_module = torch_blade.optimize(module, allow_tracing=True, model_inputs=(x, y))

# run the optimized module
blade_module(x, y)

torch_blade.optimize accepts an nn.Module object and outputs the optimized module. For more information, please refer to Quickstart for PyTorch Users.

Setup and Examples

Publications

Tutorials and Documents for Developers

Presentations and Talks

How to Contribute

Building Status

Framework Device Status
PyTorch Pre GPU pytorch_pre_gpu
PyTorch Pre CPU pytorch_pre_cpu
PyTorch2.0.0 GPU pytorch200_gpu
PyTorch2.0.0 CPU pytorch200_cpu
PyTorch2.0.0 Yitian pytorch200_yitian
PyTorch1.13.0 GPU pytorch113_gpu
PyTorch1.13.0 CPU pytorch113_cpu
PyTorch1.13.0 Yitian pytorch113_yitian
TensorFlow2.5 GPU tf250_gpu
TensorFlow2.5 CPU tf250_cpu
TensorFlow2.8 Yitian tf280_yitian

FAQ

Roadmap with mlir-hlo Project

BladeDISC is in a close relationship with mlir-hlo project. Part of the building blocks, including the MHLO Op definitions, TF to MHLO conversions, and some general purpose passes have been upstreamed to mlir-hlo repository. We'll continue to work in a close cooperative relationship with mlir-hlo project in the longer term.

Roadmap with Torch-MLIR Project

BladeDISC compiles PyTorch workloads based on Torch-MLIR. The BladeDISC Dev Team is cooperating with the community to add Torch-To-Mhlo conversion to Torch-MLIR, especially fully dynamic shape features. See RFC: llvm/torch-mlir#999. We appeal to the community developers interested in joining.

Contact Us

DingTalk