Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Can You Win Everything with A Lottery Ticket?
Authors: Tianlong Chen, Zhenyu Zhang, Jun Wu, Randy Huang, Sijia Liu, Shiyu Chang, Zhangyang Wang
TMLR 2022 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | With extensive experiments across datasets {CIFAR-10, CIFAR-100, and Image Net}, model architectures, as well as seven sparsification methods, we thoroughly characterize the trade-off between model sparsity and the all-dimension model capabilities. |
| Researcher Affiliation | Collaboration | Tianlong Chen EMAIL University of Texas at Austin Zhenyu Zhang EMAIL University of Texas at Austin Jun Wu EMAIL Amazon Web Services Randy Huang EMAIL Amazon Web Services Sijia Liu EMAIL Michigan State University MIT-IBM Watson AI Lab, IBM Research Shiyu Chang EMAIL University of California, Santa Barbara Zhangyang Wang EMAIL University of Texas at Austin |
| Pseudocode | No | The paper describes methods in regular paragraph text without structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Codes are available in https://github.com/VITA-Group/LTH-Pass. |
| Open Datasets | Yes | Experiments are conducted on CIFAR-10 (C10), CIFAR-100 (C100) (Krizhevsky & Hinton, 2009), and Image Net (IMG) (Deng et al., 2009). |
| Dataset Splits | Yes | Following the standard setup in Hendrycks & Dietterich (2019), we use the mean corruption error (m CE) to indicate model robustness to different natural corruptions... CIFAR-10/100-C and Image Net-C (Hendrycks & Dietterich, 2019) are adopted in our experiments. Following Hendrycks & Gimpel (2016); Hendrycks et al. (2018); Hein et al. (2019); Augustin et al. (2020), for CIFAR-10 experiments, CIFAR-100 (Krizhevsky & Hinton, 2009) is regarded as the Oo D dataset; for CIFAR-100 experiments, CIFAR-10 is selected as the Oo D dataset; for Image Net experiments, Image Net-O (Hendrycks et al., 2019) is the Oo D dataset. |
| Hardware Specification | No | The paper does not provide specific details regarding the hardware used for running experiments, such as exact GPU/CPU models or cloud instance specifications. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions). |
| Experiment Setup | Yes | Table 1: Details of training configurations for experiments with OMP, LTH, RP, PI approaches. Dataset Learning Rate Batch Size Epochs Optimizer Momentum Weight Decay CIFAR-10/100 0.1; 0.1 at 91,136 epoch 128 182 SGD 0.9 1 10 4 Image Net 0.4; 0.1 at 30,60,80 epoch; linearly warmup 5 epochs 1024 90 SGD 0.9 1 10 4 |