Joosung Yoon

is there any way to delete contest?

1

Hello, Nowadays, I'm working on managing customized OnlineJudge page. we're going to remove contests that were used but there are no remove button. similarly, we found some problem in contest...

- 띄어쓰기 모델 업데이트 - crf 모듈 업데이트 - https://github.com/s14t284/TorchCRF # 고려해볼 것

How can I modify theme by select box? I'm trying to apply this module. However I think, there is no event function to modify theme .. I spent with about...

Add `MegatronT5ForConditionalGeneration`

3

# What does this PR do? This PR adds the `MegatronT5ForConditionalGeneration` class, which among standard applications can be used for pretrained T5 model from NVIDIA NeMo MegatronT5 :) I also...

### Model description In NeMo Megatron, the T5 model is available, but there is currently no MegatronT5 class for huggingface, such as MegatronBERT or MegatronGPT2. I have recently finished the...

## 🚀 Feature - 기존의 mxnet등의 레거시를 제거합니다 - HF BERT로 기존의 모델을 교체합니다 - 기존에 학습한 ckpt를 새로운 모델 name에 맞게 convert합니다 ## Motivation - Repo를 공개한 이후 HF가...

text-gereneration-inference:0.9 version create `weired model.safetensors`

2

### System Info From upgrading the docker version from 0.8 to 0.9, I noticed that the GPT model started generating meaningless tokens. After conducting several tests, I discovered that the...

[QUESTION] How to set `--rotary-seq-len-interpolation-factor` for rope scaling?

2

Hi, I'm curious if it's possible to set `--rotary-seq-len-interpolation-factor` in [Megatron-LM](https://github.com/NVIDIA/Megatron-LM/blob/de4028a9d45bd65c67e1a201d9e0690bd6cb4304/megatron/arguments.py#L621) to match [huggingface's rope_scaling setting](https://huggingface.co/docs/transformers/main/model_doc/llama2#transformers.LlamaConfig.rope_scaling) (`{"type": "dynamic", "factor": 2.0}`). Is there any information you can share on how the...

AI for Social Good workshop at NeurIPS (Kim & Lim et al., 2019) PDF 링크가 public이 아닌것 같습니다

2

논문 링크 접속이 안되서 글 남깁니다.

Is it support LLM like Llama?

1

I would like to use Llama in this library with huggingface tokenizer. Is it possible? Is there a plan to support it? :) I would like to use it in...