JiHa-Kim

ai-evaluation-manifesto

create-symbolic-link

Create symbolic link

hedgehog-porcupine-softmax-mimicry

Code implementation in PyTorch of the paper "The Hedgehog & the Porcupine: Expressive Linear Attentions with Softmax Mimicry" at https://arxiv.org/abs/2402.04347

JiHa-Kim

Config files for my GitHub profile.

LLMLingua-implementation

I'm trying to get LLMLingua to work.

LLMs-self-prompt-creation

In this repository, I will keep track on an experiment where I test LLMs' abilities to craft and self-improve system prompts, and see if it leads to better results.

neural-network-from-scratch

quantize-hf-models

Quantize LLMs from HuggingFace into the GGUF format (both standard and with imatrix) using a Colab notebook.

text-generation-webui-fork

A Gradio web UI for Large Language Models. Supports transformers, GPTQ, AWQ, EXL2, llama.cpp (GGUF), Llama models.

update-llama.cpp-windows

I created a script to update llama.cpp on Windows with cmake because doing it manually is tedious.

JiHa-Kim/create-symbolic-link

Create symbolic link

JiHa-Kim/neural-network-from-scratch

JiHa-Kim/ai-evaluation-manifesto

JiHa-Kim/hedgehog-porcupine-softmax-mimicry

Code implementation in PyTorch of the paper "The Hedgehog & the Porcupine: Expressive Linear Attentions with Softmax Mimicry" at https://arxiv.org/abs/2402.04347

JiHa-Kim/JiHa-Kim

Config files for my GitHub profile.

JiHa-Kim/LLMLingua-implementation

I'm trying to get LLMLingua to work.

JiHa-Kim/LLMs-self-prompt-creation

In this repository, I will keep track on an experiment where I test LLMs' abilities to craft and self-improve system prompts, and see if it leads to better results.

JiHa-Kim/quantize-hf-models

Quantize LLMs from HuggingFace into the GGUF format (both standard and with imatrix) using a Colab notebook.

JiHa-Kim/text-generation-webui-fork

A Gradio web UI for Large Language Models. Supports transformers, GPTQ, AWQ, EXL2, llama.cpp (GGUF), Llama models.

JiHa-Kim/update-llama.cpp-windows

I created a script to update llama.cpp on Windows with cmake because doing it manually is tedious.