Hyunsoo Luke HA

icon indicating a website link https://www.linkedin.com/in/hyunsoo-ha-872aaa134/ icon indicating an email [email protected]

icon company @UpstageAI icon location Seoul, Republic Of Korea icon location AI Research Engineer at Upstage. Previously researched OCR. Currently handling various data tasks in the Data-Centric LLM team for training Upstage Solar LLM.

Results 4 issues of Hyunsoo Luke HA

Fix gfile and flags no attribute error for TensorFlow V2

There are "no attribute" errors such as `AttributeError: module 'tensorflow' has no attribute 'gfile'` `AttributeError: module 'tensorflow' has no attribute 'flags'` because tensorflow V2 is automatically installed from `pip install...

`torch.clamp` issue due to update of pytorch 1.12.0

2 comment

Since `torch.clamp` was updated in 1.12.0, the latest version of Pytorch, `torch.clamp`'s min, max argument should be loaded on same device with input tensor. https://github.com/pytorch/pytorch/pull/77035 I got an error with...

Fix torch.clamp issue #237

3 comment

This PR is related to #237 ! There are two options to fix this problem, 1. just convert `max` tensor to scalar (this PR) 2. make `max` tensor be loaded...

Validate tokenizer and model alignment before training

2 comment

First of all, thank you for creating this amazing project. I discovered a potential improvement regarding tokenizer usage while working with the codebase and have implemented the changes below. ##...

CLA Signed