# Multi-task instruction fine-tuning

Velvet-2B is a 2-billion-parameter Italian-English bilingual large language model, trained from scratch based on the Transformer architecture, focusing on text generation and comprehension tasks.

Hunyuan-7B is an open-source large language model released by Tencent. It has the ability to process 256K long texts and uses the Grouped Query Attention (GQA) mechanism, performing excellently among Chinese 7B dense models.

AceInstruct is a series of advanced SFT models, improved based on Qwen, suitable for coding, mathematics, and general tasks.

A 3B-parameter large language model based on the Llama3.2 architecture, supporting English text generation tasks, quantized using llama.cpp with imatrix

A large language model fine-tuned based on Qwen2.5-72B, specializing in text generation and instruction-following tasks

A multilingual large language model based on Qwen2.5-7B, integrating code, math, and general instruction-following capabilities

Mistral-7B instruction fine-tuned model optimized for Uzbek, supporting various NLP tasks in Uzbek and English

Phi-3 Small is a 7B-parameter lightweight cutting-edge open-source model, optimized for NVIDIA GPUs in ONNX format, supporting 8K context length with strong inference capabilities.

An 8B parameter model built on Meta's Llama-3, fine-tuned using the Bagel dataset, and adopting a single llama-3 instruction format.

An experimental fine-tuned model based on yi-34b-200k, suitable for creative writing, role-playing, and other tasks, without DPO stage applied.

A model focused on Yoruba text generation, fine-tuned based on multiple datasets.

FLAN-T5 is the instruction-fine-tuned version of the T5 model, optimized through fine-tuning on over 1,000 multilingual tasks, delivering superior performance at the same parameter count.

Platypus2-70B-instruct is a large language model based on the LLaMA 2 architecture, created by merging models from garage-bAInd and upstageAI, focusing on instruction following and logical reasoning tasks.

Manticore 13B is a large language model fine-tuned on multiple high-quality datasets based on the Llama 13B model, excelling in text generation tasks.

GPT-JT is a large language model fine-tuned based on GPT-J (6B), utilizing UL2 training objectives, excelling in classification tasks