You can deploy this model and start earning money today!

asher577/
nodesparse_pgd_20260611_l0_1.0_batchadv_actopk0.2_pertoken

Name: nodesparse_pgd_20260611_l0_1.0_batchadv_actopk0.2_pertoken
Availability: InStock

pytorch

Views

No views yet

nodesparse_pgd_20260611_l0_1.0_batchadv_actopk0.2_pertoken

Weight-sparse transformer trained with the procedure from Gao et al. (2025).

Model Details

Layers: 2
Model Dimension: 3072
Context Length: 512
Head Dimension: 16
Vocabulary Size: 4096

Sparsity

Weight Sparsity: True
Target L0 Fraction: 1
Activation Sparsity: True

Training

Dataset: jacobcd52/simplestories-tokenized
Tokenizer: SimpleStories/SimpleStories-1.25M
Total Tokens: 1,000,000,000

Training Run

W&B Run: https://wandb.ai/asher577/nodesparse_adv_pretraining/runs/m0n7jrrm

Usage

python

import torch
from huggingface_hub import hf_hub_download

# Download model
model_path = hf_hub_download(repo_id="asher577/nodesparse_pgd_20260611_l0_1.0_batchadv_actopk0.2_pertoken", filename="pytorch_model.bin")
config_path = hf_hub_download(repo_id="asher577/nodesparse_pgd_20260611_l0_1.0_batchadv_actopk0.2_pertoken", filename="config.json")

# Load (requires the SparseGPT model class from this repo)
state_dict = torch.load(model_path)