code for "AttentiveNAS Improving Neural Architecture Search via Attentive Sampling"

Last update: Oct 26, 2022

Related tags

Overview

AttentiveNAS: Improving Neural Architecture Search via Attentive Sampling

This repository contains our PyTorch training code, evaluation code and pretrained models for AttentiveNAS.

[Update 06/21] Recenty, we have improved AttentiveNAS using an adaptive knowledge distillation training strategy, see our AlphaNet repo for more details of this work. AlphaNet has been accepted by ICML'21.

[Update 07/21] We provide an example code for searching the best models of FLOPs vs. accuracy trade-offs at here.

For more details, please see AttentiveNAS: Improving Neural Architecture Search via Attentive Sampling by Dilin Wang, Meng Li, Chengyue Gong and Vikas Chandra.

If you find this repo useful in your research, please consider citing our work:

@article{wang2020attentivenas,
  title={AttentiveNAS: Improving Neural Architecture Search via Attentive Sampling},
  author={Wang, Dilin and Li, Meng and Gong, Chengyue and Chandra, Vikas},
  journal={arXiv preprint arXiv:2011.09011},
  year={2020}
}

Evaluation

To reproduce our results:

Please first download our pretrained AttentiveNAS models from a Google Drive path and put the pretrained models under your local folder ./attentive_nas_data

To evaluate our pre-trained AttentiveNAS models, from AttentiveNAS-A0 to A6, on ImageNet with a single GPU, please run:

python test_attentive_nas.py --config-file ./configs/eval_attentive_nas_models.yml --model a[0-6]

Expected results:

Name	MFLOPs	Top-1 (%)
AttentiveNAS-A0	203	77.3
AttentiveNAS-A1	279	78.4
AttentiveNAS-A2	317	78.8
AttentiveNAS-A3	357	79.1
AttentiveNAS-A4	444	79.8
AttentiveNAS-A5	491	80.1
AttentiveNAS-A6	709	80.7

Training

To train our AttentiveNAS models from scratch, please run

python train_attentive_nas.py --config-file configs/train_attentive_nas_models.yml --machine-rank ${machine_rank} --num-machines ${num_machines} --dist-url ${dist_url}

We adopt SGD training on 64 GPUs. The mini-batch size is 32 per GPU; all training hyper-parameters are specified in train_attentive_nas_models.yml.

Additional data

A (sub-network config, FLOPs) lookup table could be used for constructing the architecture distribution under FLOPs-constraints.
A accuracy predictor trained via scikit-learn, which takes a subnetwork configuration as input, and outputs its predicted accuracy on ImageNet.
- Convert a subnetwork configuration to our accuracy predictor compatibale inputs:
```
    res = [cfg['resolution']]
    for k in ['width', 'depth', 'kernel_size', 'expand_ratio']:
        res += cfg[k]
    input = np.asarray(res).reshape((1, -1))
```

License

The majority of AttentiveNAS is licensed under CC-BY-NC, however portions of the project are available under separate license terms: Once For All is licensed under the Apache 2.0 license.

Contributing

We actively welcome your pull requests! Please see CONTRIBUTING and CODE_OF_CONDUCT for more info.

Comments

Acc predictor

The following is how to convert a subnetwork configuration to accuracy predictor compatibale inputs as u provide" res = [cfg['resolution']] for k in ['width', 'depth', 'kernel_size', 'expand_ratio']: res += cfg[k] input = np.asarray(res).reshape((1, -1))

Does the order ['resolution', 'width', 'depth', 'kernel_size', 'expand_ratio'] matter?

opened by Tongzhou0101 3
Accuracy Predictor

Hi, thanks for the great work! I have a question about the usage of the accuracy predictor.

Specifically, a predictor is used to get the acc of sub-networks and rank them during training, as described in your paper. But in the code, I didn't find where the predictor is used, like here (https://github.com/facebookresearch/AttentiveNAS/blob/88ad92f82dc343a0e7d681f1fb9a8deeb45be928/train_attentive_nas.py#L291), the criterion(model(input)) is used the get the predicted acc instead.

I am a little confused about this part, is there any important code I missed or any statement I misunderstood? Looking forward to your reply : )

opened by minghaoBD 2
The actual way to get best/worst pareto models

It seems that the accuracy predictor is not fed into the sampler. Instead of using accuracy predictor, the code shows the way to get the pareto models is to rank the computed losses of k models after running forward. I am confused that the way to get the best/worst pareto model during training is different from the details in the paper. Am I misunderstanding the paper or missing the details of the code?

opened by RachelXu7 1
The supernet appears to be reinitialize during the training process

The supernet appears to be reinitialize during the training process. I have met this issue when I'm running the AlphaNet. The log is as follows: Example 1: [10/09 16:00:53]: Epoch: [4][ 50/312] Time 2.075 ( 2.485) Data 0.000 ( 0.273) Loss 4.9844e+00 (4.9407e+00) Acc@1 17.43 ( 16.29) Acc@5 37.01 ( 35.80) [10/09 16:01:15]: Epoch: [4][ 60/312] Time 2.258 ( 2.431) Data 0.000 ( 0.228) Loss 4.9118e+00 (4.9424e+00) Acc@1 15.50 ( 16.19) Acc@5 34.94 ( 35.68) [10/09 16:01:37]: Epoch: [4][ 70/312] Time 2.368 ( 2.400) Data 0.000 ( 0.196) Loss 6.8941e+00 (5.1301e+00) Acc@1 0.10 ( 14.50) Acc@5 0.81 ( 32.05) [10/09 16:01:59]: Epoch: [4][ 80/312] Time 1.940 ( 2.374) Data 0.000 ( 0.172) Loss 6.8695e+00 (5.3466e+00) Acc@1 0.10 ( 12.73) Acc@5 0.76 ( 28.20)

Example 2: [10/11 08:46:30]: Epoch: [169][170/312] Time 2.279 ( 2.272) Data 0.000 ( 0.082) Loss 3.7633e+00 (3.6145e+00) Acc@1 41.94 ( 43.52) Acc@5 64.28 ( 67.07) [10/11 08:46:53]: Epoch: [169][180/312] Time 2.159 ( 2.270) Data 0.000 ( 0.077) Loss 3.7879e+00 (3.6247e+00) Acc@1 39.58 ( 43.30) Acc@5 63.65 ( 66.86) [10/11 08:47:15]: Epoch: [169][190/312] Time 2.206 ( 2.266) Data 0.000 ( 0.073) Loss 6.7652e+00 (3.6773e+00) Acc@1 0.22 ( 42.50) Acc@5 0.68 ( 65.76) [10/11 08:47:37]: Epoch: [169][200/312] Time 2.339 ( 2.262) Data 0.000 ( 0.069) Loss 6.8340e+00 (3.8188e+00) Acc@1 0.07 ( 40.39) Acc@5 0.44 ( 62.51)

After re-initialization, the supernet gradually fits again if training continues. Is it because of the sandwich rule？

opened by liwei9719 1
Codes about evolutionary search on the ImageNet validation set

It seems that the codes about evolutionary search are not included in this repo, will you guys open source them?

Also, the codes for generating the look-up table are also not included, it will be useful to have them:)

opened by chenbohua3 1
Question about ImageNet dataset

Could you please clarify the exact ImageNet Dataset you have used for training and testing? Is it the original ILSVRC2012 from https://image-net.org/challenges/LSVRC/2012/2012-downloads.php or from somewhere else? Is it train and validation split or you have used test split as well? What was the number of images in train and validation dataset?

Also have you made any preprocessing, e.g. resizing or you took the raw images?

opened by marianpetruk 1
pretrained AttentiveNAS models I downloaded is corrupt

pretrained AttentiveNAS models I downloaded from https://drive.google.com/file/d/1cCla-OQNIAn-rjsY2b832DuP59ZKr8uh/view?usp=sharing is corrupt.I don't know if I'm wrong. thanks

opened by yangyang90 2

Owner

Facebook Research

GitHub

TensorFlow code for the neural network presented in the paper: "Structural Language Models of Code" (ICML'2020)

SLM: Structural Language Models of Code This is an official implementation of the model described in: "Structural Language Models of Code" [PDF] To ap

73 Nov 6, 2022

Inference code for "StylePeople: A Generative Model of Fullbody Human Avatars" paper. This code is for the part of the paper describing video-based avatars.

NeuralTextures This is repository with inference code for paper "StylePeople: A Generative Model of Fullbody Human Avatars" (CVPR21). This code is for

Visual Understanding Lab @ Samsung AI Center Moscow

18 Oct 6, 2022

A code generator from ONNX to PyTorch code

onnx-pytorch Generating pytorch code from ONNX. Currently support onnx==1.9.0 and torch==1.8.1. Installation From PyPI pip install onnx-pytorch From

94 Jan 6, 2023

This is the code for our KILT leaderboard submission to the T-REx and zsRE tasks. It includes code for training a DPR model then continuing training with RAG.

KGI (Knowledge Graph Induction) for slot filling This is the code for our KILT leaderboard submission to the T-REx and zsRE tasks. It includes code fo

72 Jan 6, 2023

Convert Python 3 code to CUDA code.

Py2CUDA Convert python code to CUDA. Usage To convert a python file say named py_file.py to CUDA, run python generate_cuda.py --file py_file.py --arch

3 Jul 14, 2021

Empirical Study of Transformers for Source Code & A Simple Approach for Handling Out-of-Vocabulary Identifiers in Deep Learning for Source Code

Transformers for variable misuse, function naming and code completion tasks The official PyTorch implementation of: Empirical Study of Transformers fo

56 Nov 15, 2022

Reference implementation of code generation projects from Facebook AI Research. General toolkit to apply machine learning to code, from dataset creation to model training and evaluation. Comes with pretrained models.

This repository is a toolkit to do machine learning for programming languages. It implements tokenization, dataset preprocessing, model training and m

408 Jan 1, 2023

Code for the prototype tool in our paper "CoProtector: Protect Open-Source Code against Unauthorized Training Usage with Data Poisoning".

CoProtector Code for the prototype tool in our paper "CoProtector: Protect Open-Source Code against Unauthorized Training Usage with Data Poisoning".

1 Oct 26, 2021

Low-code/No-code approach for deep learning inference on devices

EzEdgeAI A concept project that uses a low-code/no-code approach to implement deep learning inference on devices. It provides a componentized framewor

7 Apr 5, 2022

Code for all the Advent of Code'21 challenges mostly written in python

Advent of Code 21 Code for all the Advent of Code'21 challenges mostly written in python. They are not necessarily the best or fastest solutions but j

4 May 26, 2022

Code to use Augmented Shapiro Wilks Stopping, as well as code for the paper "Statistically Signifigant Stopping of Neural Network Training"

This codebase is being actively maintained, please create and issue if you have issues using it Basics All data files are included under losses and ea

32 Nov 9, 2021

Opinionated code formatter, just like Python's black code formatter but for Beancount

beancount-black Opinionated code formatter, just like Python's black code formatter but for Beancount Try it out online here Features MIT licensed - b

16 Oct 11, 2022

a delightful machine learning tool that allows you to train, test and use models without writing code

igel A delightful machine learning tool that allows you to train/fit, test and use models without writing code Note I'm also working on a GUI desktop

3k Jan 5, 2023

Pytorch Lightning code guideline for conferences

Deep learning project seed Use this seed to start new deep learning / ML projects. Built in setup.py Built in requirements Examples with MNIST Badges

1k Jan 2, 2023

Automatically Build Multiple ML Models with a Single Line of Code. Created by Ram Seshadri. Collaborators Welcome. Permission Granted upon Request.

Auto-ViML Automatically Build Variant Interpretable ML models fast! Auto_ViML is pronounced "auto vimal" (autovimal logo created by Sanket Ghanmare) N

397 Dec 30, 2022

Applications using the GTN library and code to reproduce experiments in "Differentiable Weighted Finite-State Transducers"

gtn_applications An applications library using GTN. Current examples include: Offline handwriting recognition Automatic speech recognition Installing

68 Dec 29, 2022