Skip to content

Instantly share code, notes, and snippets.

@racinmat
Last active November 28, 2024 07:50
Show Gist options
  • Select an option

  • Save racinmat/413ad52e453e738dde7b8936f63a7f56 to your computer and use it in GitHub Desktop.

Select an option

Save racinmat/413ad52e453e738dde7b8936f63a7f56 to your computer and use it in GitHub Desktop.
Odkazy z přednášky pro DataScript - Jak hostovat nejen jazykové modely v produkci

Prezentace na google drive

https://docs.google.com/presentation/d/1-zkGvh-RBqf-h7ViEArsnRvKnjxtgxI4/edit?usp=sharing&ouid=102632222370728552977&rtpof=true&sd=true

Zdroje a materiály

awesome llms https://github.com/Hannibal046/Awesome-LLM

Deprekování modelů

openai embeddingy

Embeddingy a GDPR

jednotlivé LLMs

repozitáře

Docker

Docker desktop alternativy

Conda

anaconda license

cat >$CONDA_ROOT/.condarc <<EOL
default_channels:
  - conda-forge
  - pytorch
  - nvidia
  - rapidsai
channels:
  - defaults
allow_other_channels: False
EOL

za proxy

cat >$CONDA_ROOT/.condarc <<EOL
default_channels:
  - https://<proxy>/anaconda.org-conda-forge/
  - https://<proxy>/anaconda.org-pytorch/
  - https://<proxy>/anaconda.org-nvidia/
  - https://<proxy>/anaconda.org-rapidsai/
channels:
  - defaults
allow_other_channels: False
EOL

nastavení pipu:

obsah pip.conf

[global]
index-url=https://<proxy for pypi>/simple
extra-index-url=https://<local pypi repository>/simple

frameworky

https://www.assemblyai.com/blog/pytorch-vs-tensorflow-in-2023/

formáty

backendy, knihovny, frameworky

ONNX

GGUF a GGML

srovnání https://betterprogramming.pub/frameworks-for-serving-llms-60b7f7b23407 Speeding up inference https://betterprogramming.pub/speed-up-llm-inference-83653aa24c47 TogetherAI inference server: muc faster than TGI or vllm: https://medium.com/@datadrifters/the-worlds-fastest-llm-inference-engine-3x-faster-than-vllm-and-tgi-a2ed9e33c55f

kvantizace

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment