has-8

Hybrid Layer-Wise ANN-SNN With Surrogate Spike Encoding-Decoding Structure

Quickstart

This is a quickstart guide on how to use our model as a package

Step 1: Environment setup and repo download

To setup the environment testing with this encoder, you will need Pytorch and SpikingJelly. We suggest using conda environment with:

$ conda create -n env python=3.12.2
$ conda install pytorch=2.3.0 torchvision=0.18.0 torchaudio=2.3.0 pytorch-cuda=12.1 -c pytorch -c nvidia #As latest pytorch conda guide, change cuda version suitable to your case.
$ pip install spikingjelly
$ pip install git+https://github.com/luutn2002/has-8.git

or clone and modify locally:

$ git clone https://github.com/luutn2002/has-8.git
$ cd has-8
$ pip install -r requirements.txt

Step 2: Usage

To ensure reproducibility, remember to use static random seed:

import torch
import numpy as np
import random

seed = 3407

torch.manual_seed(seed)
random.seed(seed)
np.random.seed(seed)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False

To use the HAS-8-VGG model, we can simply import as a normal Pytorch model:

from has_8 import has8_vgg_b16_m2_d4, has8_vgg_b24_m2_d4

IN_CHANNELS = 3
NUM_CLASSES = 10
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

model = has8_vgg_b16_m2_d4(in_channels=IN_CHANNELS, 
                           out_channels=NUM_CLASSES).to(DEVICE)
# Or
model = has8_vgg_b24_m2_d4(input_channels=IN_CHANNELS, 
                           out_channels=NUM_CLASSES).to(DEVICE)

For HAS-8-ResNet:

from has_8 import has8_rn_b32_m2_d4, has8_rn_b64_m2_d4

IN_CHANNELS = 3
NUM_CLASSES = 10
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

model = has8_rn_b32_m2_d4(input_channels=IN_CHANNELS, 
                          num_classes=NUM_CLASSES).to(DEVICE)
# Or
model = has8_rn_b64_m2_d4(input_channels=IN_CHANNELS, 
                          num_classes=NUM_CLASSES).to(DEVICE)

Preprocessing

All used dataset and transform is included in torchvision. Preprocess with torch for images larger than 224x224:

BATCH_SIZE = 16
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

TRANSFORM_TRAIN = torchvision.transforms.Compose([
    torchvision.transforms.RandomHorizontalFlip(),
    torchvision.transforms.Resize(224),
    torchvision.transforms.ToTensor(),
])

TRANSFORM_TEST = torchvision.transforms.Compose([
    torchvision.transforms.Resize(224),
    torchvision.transforms.ToTensor(),
])

train_set = ...

val_set = ...

train_loader = torch.utils.data.DataLoader(train_set, 
                                            batch_size=BATCH_SIZE, 
                                            shuffle=True,
                                            #generator=torch.Generator(device=DEVICE),
                                            collate_fn=lambda x: tuple(x_.to(DEVICE) for x_ in torch.utils.data.dataloader.default_collate(x)))
test_loader = torch.utils.data.DataLoader(val_set, 
                                            batch_size=BATCH_SIZE,
                                            shuffle=True,
                                            #generator=torch.Generator(device=DEVICE),
                                            collate_fn=lambda x: tuple(x_.to(DEVICE) for x_ in torch.utils.data.dataloader.default_collate(x)))

Optimizer settings

For Adam, used in small scale datasets:

optimizer = torch.optim.Adam(model.parameters(), 
                                 lr=1e-3, 
                                 weight_decay=1e-3)

For SGDR, used in ImageNet:

BATCH_SIZE = 16
EPOCHS = 100

base_lr = 0.1*(BATCH_SIZE/256)
optimizer = torch.optim.SGD(
    model.parameters(),
    lr=base_lr,
    momentum=0.9,
    weight_decay=5e-4,
    nesterov=True
)

warmup_scheduler = torch.optim.lr_scheduler.LinearLR(optimizer, start_factor=base_lr/10, total_iters=5)
cosine_scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=EPOCHS - 5)
scheduler = torch.optim.lr_scheduler.SequentialLR(optimizer, schedulers=[warmup_scheduler, cosine_scheduler], milestones=[5])

License

Source code is licensed under MIT License.

Contribution guidelines

Please open an issue or pull request if there are bugs or contribution to be made. Thank you.

Citations

Paper is under review. Temporarily please cite as:

@article{luu2025hybrid,
  title={Hybrid Layer-Wise ANN-SNN With Surrogate Spike Encoding-Decoding Structure},
  author={Luu, Nhan T and Luu, Duong T and Pham, Nam N and Truong, Thang C},
  journal={arXiv preprint arXiv:2509.24411},
  year={2025}
}