lrs33 / ragideal-chat

RAGIdeal-Chat LLM

14B

5 Pulls Updated 4 days ago

4 days ago

521824258585 · 11GB

model
qwen2
·
14.8B
·
Q5_K_M
template
{{ if .System }}<|im_start|>system {{ .System }}<|im_end|>{{ end }}<|im_start|>user {{ .Prompt }}<|im_end|> <|im_start|>assistant
params
{"num_ctx":4096,"stop":["<|im_start|>","<|im_end|>"],"temperature":0.05}

Readme

RAGIdeal-Chat is a retrieval-augmented generation (RAG) model specifically tailored for Chinese environments. This model not only utilizes the Chinese translation of the GlaiveAI/RAG-v1 dataset but also integrates query rewriting datasets, aiming to significantly enhance its expressiveness and practicality in Chinese application scenarios. By implementing low-rank adaptation (LoRA) fine-tuning using the Qwen2.5-14B-Instruct model, along with the integration of adapter models, and processing steps such as quantization using llama.cpp, RAGIdeal-Chat is optimized to deliver superior performance.