일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | ||||
4 | 5 | 6 | 7 | 8 | 9 | 10 |
11 | 12 | 13 | 14 | 15 | 16 | 17 |
18 | 19 | 20 | 21 | 22 | 23 | 24 |
25 | 26 | 27 | 28 | 29 | 30 | 31 |
- ubuntu
- instruction tuning
- Lora
- Mac
- backbone
- ViT
- llama-4-scout-17b-16e-instruct
- gemma-3-27b-it
- transformer
- CPT
- nccl
- prompt
- diffusion
- sfttrainer
- tensor-parallel
- error: mkl-service + intel(r)
- multi-gpu
- vLLM
- Text-to-Image
- llm
- langchain
- PEFT
- gemma-3
- Fine-tuning
- lora+
- Gemma
- domain-adapted pre-training
- llama-4
- gemma2
- glibcxx
- Today
- Total
목록전체 글 (32)
꾸준하게
적용 방법1. vi ~/.bashrc에 아래 내용 삽입 후2. source ~/.bashrc 명령어를 통해 적용# enable current branchparse_git_branch() { git branch 2> /dev/null | sed -e '/^[^*]/d' -e 's/* \(.*\)/(\1)/'}export PS1="\e[01;32m\u@\h \[\e[34m\]\w\[\e[91m\]\$(parse_git_branch)\[\e[00m\]$ "

발생 이슈NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running. 해결방법 : Azure가 아닌 일반 local machine에서 해결 가능한 방법Secure Boot : Enable 상태일 경우 Bios 모드에 진입해서 Disable로 바꾸어야한다. Secure Mode일때 Nvidia driver를 설치할 때 Sign을 막아서 생기는 문제로 Disable로 수정하면 해결이 된다. 만약 Azure를 사용중이라면 VM 생성 시 Security type을 Trusted launch virtual machin..

LangChain AgentOllama를 통해 gemma-2-9b 모델을 띄우고, fastAPI를 통해 간단한 엔드포인트를 만들었다. 터미널 출력을 보니 LLM 모델 혼자서 사람처럼 추론하는 모습이 신기해서 작동과정에 궁금증이 생겼다. 우선, 이를 Python상에서 디버깅해보니 다음과 같이 프롬프트 형식으로 내용이 들어간다. agent.agent.llm_chain.prompt.template 'Answer the following questions as best you can. You have access to the following tools:\n\nget_today(input_text: str) -> str - 오늘 날짜를 반환합다.\n\nUse the following format:\n\nQues..
.vscode/launch.json 에서 "justMyCode": false 를 추가해주면 해결이 된다.{ "version": "0.2.0", "configurations": [ { "name": "Python: Debug Current File", "type": "python", "request": "launch", "program": "${file}", "console": "integratedTerminal", "stopOnEntry": true, "justMyCode": false }, ]}
아래 코드를 실행하니 위와 같은 오류가 발생하여 brew install ghostscript 명령어로 설치를 해주었다.하지만 설치를 했음에도 같은 오류가 발생. Stackoverflow에서 해결방법을 얻게되어 글을 작성하였다. https://stackoverflow.com/questions/75391805/need-to-install-ghostscript-to-mac-path 실행 명령어from llama_index.readers.pdf_table import PDFTableReaderfrom pathlib import Pathreader = PDFTableReader()pdf_path = Path(file_path)documents = reader.load_data(file=pdf_path, pages..
증상Mac 환경에서 conda 설치 후, pip install transformers 실행 시 나타난 오류이다.(llm) ➜ ~ python -m pip install --upgrade piperror: externally-managed-environment× This environment is externally managed╰─> To install Python packages system-wide, try brew install xyz, where xyz is the package you are trying to install. If you wish to install a Python library that isn't in Homebrew, use a virtual envir..

arXiv 2023 [paper] [code] [sample]Flavio Schneider, Ojasv Kamal, Zhijing Jin, Bernhard SchölkopfETH Zürich | IIT Kharagpur | MPI for Intelligent Systems & ETH Zürich | MPI for Intelligent Systemssubmitted on 27 Jan 2023 (v1)last revised 23 Oct 2023 (v3) Abstract최근 텍스트 분야의 거대 생성모델이 급격히 개발이 되었지만 텍스트와 음악을 연결 짓는 연구가 아직은 부족한 편이다. 저자는 텍스트를 입력으로 하여 긴 길이의 음악도 생성할 수 있는 text-to-music 생성 모델 Moûsai을 제안하였다. ..

CVPR 2023 [paper] [code]Wenhai Wang, Jifeng Dai, Zhe Chen, Zhenhang Huang, Zhiqi Li, Xizhou Zhu, Xiaowei Hu, Tong Lu, Lewei Lu, Hongsheng Li, Xiaogang Wang, Yu QiaoShanghai AI Laboratory | Tsinghua University | Nanjing University | SenseTime Research | The Chinese University of Hong KongSubmitted on 10 Nov 2022last revised 17 Apr 2023 (v4) Abstract저자는 근래 나온 ViT 모델과 달리 CNN은 아직 초기 상태에 머물러 있다고 주장하며..