# English Generation

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

A quantized version based on Gemma 2.9B, optimized using llama.cpp, suitable for running in LM Studio.

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

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

Meta-Llama-3-70B-Instruct is a 70B parameter-scale instruction-tuned large language model released by Meta, optimized for dialogue use cases, with outstanding performance in safety and helpfulness.

This is the GGUF quantized version of the 8 billion parameter instruction-tuned model from the Meta Llama 3 series, optimized for dialogue scenarios and demonstrating excellent performance in multiple benchmark tests.

Llama 2 is a 70-billion-parameter conversation-optimized large language model developed by Meta, surpassing most open-source dialogue models in benchmark tests with safety comparable to mainstream proprietary models