asgeirtj/system_prompts_leaks
This repository provides a collection of extracted system prompts from various large language models (LLMs) including Anthropic's Claude, OpenAI's ChatGPT, Google's Gemini, and xAI's Grok, updated ...
Awesome AI Agents › Prompt Libraries
arcee-ai/mergekit
mergekit is a toolkit designed for merging pre-trained large language models efficiently, even in resource-constrained environments. It employs an out-of-core approach that allows merging to be performed entirely on CPU or with minimal GPU resources, such as 8 GB of VRAM. The toolkit supports a variety of merging algorithms and is compatible with popular models like Llama, Mistral, GPT-NeoX, and StableLM. The main advantage of merging models using mergekit is that it combines the strengths of different models directly in the weight space, avoiding the computational overhead of ensembling multiple models or the need for additional training. This results in a single versatile model that maintains the inference cost of one model while potentially improving performance and capabilities. Key features of mergekit include support for multiple merge methods, lazy loading of tensors to reduce memory usage, interpolated gradients for parameter values, piecewise assembly of models from layers (referred to as "Frankenmerging"), mixture of experts merging, LoRA extraction, and evolutionary merge methods. The toolkit also offers a graphical user interface backed by mega-GPU resources through the Arcee App, making the merging process more accessible. Users can configure merges using YAML files specifying models, slices, parameters, tokenizer settings, and chat templates. The toolkit facilitates uploading merged models to the Hugging Face Hub, including generating a README.md for model cards. mergekit is particularly useful for researchers and developers who want to combine specialized models into a single model, transfer capabilities without retraining, optimize trade-offs between model behaviors, and create new capabilities through creative combinations. It supports both CPU and GPU execution, making it flexible for different hardware setups. The project emphasizes ease of use, configurability, and efficient resource utilization, making it a valuable tool in the large language model ecosystem.
https://github.com/arcee-ai/mergekit
Also in Prompt Libraries
HKUDS/RAG-Anything
RAG-Anything is a next-generation all-in-one multimodal Retrieval-Augmented Generation system that processes and queries diverse document content including text, images, tables, and equations within a unified framework.
dzhng/deep-research
Open Deep Research is an AI-powered research assistant that performs iterative, deep research on any topic by combining search engines, web scraping, and large language models to generate comprehensive markdown reports.
dottxt-ai/outlines
Outlines is an open-source tool that enables structured text generation with large language models, ensuring predictable and schema-compliant outputs for various applications.
microsoft/RD-Agent
RD-Agent is an open-source R&D automation tool by Microsoft designed to automate and enhance industrial research and development processes focused on data and models using AI.
icip-cas/PPTAgent
PPTAgent is an innovative system that automatically generates and evaluates high-quality presentations from documents using a two-phase approach and a comprehensive multi-dimensional evaluation framework.
serenakeyitan/awesome-notebookLM-prompts
A curated collection of NotebookLM and Kael.im slide prompts designed to generate presentation-ready decks from various content types.
snap-stanford/Biomni
Biomni is a general-purpose biomedical AI agent that autonomously executes diverse research tasks across multiple biomedical subfields by integrating advanced LLM reasoning, retrieval-augmented planning, and code-based execution.