# Instruction Optimization

Qwen2.5-Coder-7B-Instruct is a 7B-parameter code generation model based on the Qwen2.5 architecture, specializing in instruction fine-tuning, suitable for code generation and programming assistance tasks.

A 3B parameter instruction fine-tuned model based on Qwen2.5 architecture, optimized for training speed using Unsloth and Huggingface TRL library

A merged model based on Meta-Llama-3.1-8B-Instruct, fine-tuned for legal summarization tasks, capable of converting legal terminology into concise and understandable summaries.

Fine-tuned version based on Qwen2.5-VL-3B model, trained using Unsloth and Huggingface TRL library, achieving 2x inference speed improvement

An enhanced version of the Qwen2.5-14B model fused using the TIES method, focusing on instruction following and dialogue optimization

The fifth-generation Qwen2.5-YOYO model integrates features from multiple advanced models, optimizes the model merging formula, and supports a context length of 1 million tokens.

A dialogue model based on the ChatML architecture, generated by merging pre-trained language models using the mergekit tool

Multilingual E5 large instruction model supporting text embedding and classification tasks in multiple languages

Multilingual E5 large instruction model, supporting multiple languages and tasks, including classification, retrieval, clustering, etc.

Multilingual E5 large instruction model, supporting multiple languages and tasks including classification, retrieval, clustering, etc.

Multilingual E5 large instruction model supporting text embedding and classification tasks in multiple languages

Llama 3.3 70B Instruct AWQ INT4 is the 4-bit quantized version of the Meta Llama 3.3 70B Instruct model, optimized for multilingual dialogue use cases and text generation tasks.

A custom large language model by NVIDIA, designed to enhance the usefulness of responses generated by LLMs to user queries.

4-bit quantized version based on Meta-Llama-3.1-8B-Instruct, efficiently trained using Unsloth and TRL libraries

A reward model fine-tuned based on Qwen/Qwen2-0.5B-Instruct, used to evaluate and optimize the quality of generated content

Llama-3.1-Storm-8B is an improved model based on Llama-3.1-8B-Instruct, demonstrating excellent performance in multiple benchmarks and is suitable for dialogue and function calling tasks.

The multilingual e5 series is one of the best choices for multilingual embedding models, supporting multiple languages and tasks.

A 4-bit quantized version based on Meta-Llama-3.1-8B-Instruct, trained using Unsloth and Huggingface TRL libraries, achieving 2x speed improvement.

This is an improved version of the Llama-3.1-8B-Instruct model with refusal behavior eliminated via LoRA technology

INT4 quantized version of Llama 3.1 8B Instruct, quantized using AutoAWQ tool, suitable for multilingual dialogue scenarios.

Quantized version of Mistral-7B-Instruct-v0.3, offering multiple quantization options to accommodate different hardware requirements

Phi-3-Mini-4K-Instruct is a 3.8 billion parameter lightweight cutting-edge open-source model trained on the Phi-3 dataset, emphasizing high quality and inference-intensive characteristics.

Suzume 8B is a multilingual fine-tuned version based on Llama 3, trained on nearly 90,000 multilingual dialogues to enhance multilingual communication capabilities while maintaining Llama 3's intelligence level.

ECE-TW3-JRGL-V5 is a new model obtained by merging the MoMo-72B-lora-1.8.7-DPO and alpaca-dragon-72b-v1 models through mergekit, integrating the advantages of multiple models.

This is a GQL language model fine-tuned on CodeLlama-7b-Instruct-hf, specifically designed for the TuGraph Analytics graph computing system.

A version fine-tuned using the IFprompMKR dataset with qlora on the microsoft/phi-1_5 model, primarily for text generation tasks.