Compare commits

..

4 commits

4 changed files with 77 additions and 4 deletions

View file

@ -145,6 +145,7 @@ const today = new Date();
color: rgb(var(--gray)); color: rgb(var(--gray));
text-align: center; text-align: center;
margin-top: 60px; margin-top: 60px;
font-size: 16px;
} }
.social-links { .social-links {
display: flex; display: flex;

View file

@ -0,0 +1,73 @@
---
title: 'Local Llama Quickstart'
description: 'A collection of guides to get started with local llamas'
updateDate: 'Dec 31 2023'
heroImage: '/images/llama/llama-cool.avif'
---
<p style="font-size: max(2vh, 10px); margin-top: 0; text-align: right">
Midjourney-generated llama from <a href="https://pub.towardsai.net/meet-vicuna-the-latest-metas-llama-model-that-matches-chatgpt-performance-e23b2fc67e6b">Medium</a>
</p>
"Llama" refers to a Large Language Model (LLM). "Local llama" refers to a
locally-hosted (typically open source) llama, in contrast to commercially hosted
ones.
# Local LLaMa Quickstart
I've recently become aware of the open source LLM (local llama) movement. Unlike
traditional open source, the speed at which this field is moving is
unprecedented. Several breakthroughs come out on a weekly basis and information
dated further back than a month is often functionally deprecated.
This collection was gathered in late December 2023, with the intent to help
anyone looking to get caught up with the field.
## Models
#### Model Sources
- [Pre-quantized models](https://huggingface.co/TheBloke)
## Practical use
#### Interfaces
- [List of web UIs](https://www.reddit.com/r/LocalLLaMA/comments/1847qt6/llm_webui_recommendations)
- [SillyTavern tutori]()
- https://www.reddit.com/r/CharacterAi_NSFW/comments/13yy8m9/indepth_explanation_on_how_to_install_silly/ :: tutorial sillytavern webui
- [Jailbreaking GPT when prompting for characters](https://rentry.org/GPTJailbreakPrompting)
- [Guidelines for prompting for characters](https://rentry.org/NG_CharCard)
- [ChatML from OpenAI is quickly becoming the standard for
prompting](https://news.ycombinator.com/item?id=34988748)
#### Training
- [Teaching llama a new language through tuning](https://www.reddit.com/r/LocalLLaMA/comments/18oc1yc/i_tried_to_teach_mistral_7b_a_new_language)
- [Mergekit - MoE training framework](https://github.com/cg123/mergekit)
- [Axolotl - Fine tuning framework](https://github.com/OpenAccess-AI-Collective/axolotl)
- [Unsloth - Fine tuning accelerator](https://github.com/unslothai/unsloth)
- llama.cpp vs Transformers vs LangChain vs PyTorch :: library framework tool standard
#### Tutorials
- [Karpathy builds and trains a llama](https://www.youtube.com/watch?v=kCc8FmEb1nY)
- [Build a llama DIY by freecodecamp](https://www.youtube.com/watch?v=UU1WVnMk4E8)
- [Understanding different quantization methods](https://www.youtube.com/watch?v=mNE_d-C82lI)
#### Servers
- [Ollama.ai](https://ollama.ai/)
- [llama.cpp](https://github.com/ggerganov/llama.cpp)
## Hardware
#### GPU stuff
- [Choosing a GPU for deeplearning](https://timdettmers.com/2023/01/30/which-gpu-for-deep-learning/)
#### Cloud renting
- [Kaggle](https://kaggle.com) - 30h/week free, enough VRAM for 7B models
- [Lambda Labs](https://lambdalabs.com) - Huge instances, competitive pricing
- [Runpod](https://runpod.io) - Bad pricing, community cloud option
- [Paperspace](https://paperspace.com) - Requires subscription, terrible pricing
- [Genesis Cloud](https://genesiscloud.com) - Reddit says it's affordable... I can't verify
- [Vast.ai](https://vast.ai) - Very affordable, especially the "interruptible" ones
## Research and Other Blogs
- [Mixture of Experts explained](https://goddard.blog/posts/clown-moe/)
- [A summary of the local llama scene in December 2023](https://www.reddit.com/r/LocalLLaMA/comments/18mwd6j/how_is_the_scene_currently/)

View file

@ -2,12 +2,11 @@
title: 'Llama Quantization Methods' title: 'Llama Quantization Methods'
description: 'A short overview of modern quantization methods in language models' description: 'A short overview of modern quantization methods in language models'
updateDate: 'Dec 31 2023' updateDate: 'Dec 31 2023'
heroImage: '/images/tiny-llama-logo.avif' heroImage: '/images/llama/pink-llama.avif'
--- ---
<p style="font-size: max(2vh, 10px); margin-top: 0; text-align: right"> <p style="font-size: max(2vh, 10px); margin-top: 0; text-align: right">
"TinyLlama logo" by <a href="https://github.com/jzhang38/TinyLlama">The Midjourney-generated llama
TinyLlama project</a>. Licensed under Apache 2.0
</p> </p>
"Llama" refers to a Large Language Model (LLM). "Local llama" refers to a "Llama" refers to a Large Language Model (LLM). "Local llama" refers to a

View file

@ -54,7 +54,7 @@ const posts = (await getCollection('llama')).sort(
width={720} width={720}
height={360} height={360}
class="rounded-2xl grayscale" class="rounded-2xl grayscale"
src={ post.data.heroImage ? post.data.heroImage : "/terminal.webp" } src={ post.data.heroImage ? post.data.heroImage : "/images/tiny-llama-logo.avif" }
alt="" /> alt="" />
<h4 <h4
class="title text-xl"> class="title text-xl">