VQGraph

VQGraph copied to clipboard

Questionable distillation technique

1

I noticed in Table 12 of your paper, the hyperparameter ($\beta$) is set to a very low value, 1e-8, which suggests that the proposed code-based distillation process plays an almost...

andyperio

您好，请问一下，您使用余弦相似度在蒸馏的时候tea_logit的分布是不是过于光滑了？几乎都是0.0002和0.0003。以及蒸馏时学生模型使用L2计算，与余弦不太符合。还有使用logit_y=True是为什么？放缩吗？也不对吧

gzoftju

Dear authors, Thank you for releasing your work! During my learning, I find that an undefined function "`rand_train_test_idx()`" is invoked in line 284 of the `dataloader.py`: ```py def get_idx_split(self, split_type="random",...

SwaggyZhang

The date of graph tokenizer training release

4

Hi, Thank you for your work, very exciting! Is there a planned date to release the code graph tokenizer training? Thank you!

RedekopEP

graph tokenizer training细节了解

2

您好，看到你的工作很感兴趣，想进一步了解一下graph tokenizer训练细节，期待您的开源！

guoswang

Missing the code of graph tokenizer training

2

Hi, Thank you for your code. Could you please relase the code graph tokenizer training ?

LLMforScience

It did not converge on the ogbn-arxiv dataset.

1

When running on ogbn-arxiv, we found the following error. ```python 20240528-09:53:39: output_dir: /public/home/jialh/metaHiC/models/01VQGraph/outputs/transductive/ogbn-arxiv/SAGE/seed_0 20240528-09:53:39: Total 169343 nodes. 20240528-09:53:39: Total 2501829 edges. 20240528-09:53:39: Load data with max_memory_allocated: 0.0000Gb | max_memory_cached: 0.0000Gb...

JiaLonghao1997

Hi, thanks for your wonderful work! This work enlights me a lot, but there are some questions I am confused about: 1. from your source code, whether the **node representation**...

ChloeKiwi

Hi authors, Thank you for the great paper and code. I am currently trying to reproduce the results of VQGraph. I noticed that when training the Graph Tokenizer (the SAGE...

Anyna918