Large Lanuage Models

An initial exploration of Large Language Models, The cover image was created by DALL·E

The advent of large language models(LLM) has profoundly captured my interest. 
Well after ending my internship at 4th Paradigm, almost half a year later, 
my former mentor invited me to explore applications on LLM with him. 
Having gained a preliminary understanding of multimodal cellular data, 
I've recognized the substantial potential of LLM in contextual processing 
for such complex information. For this purpose, I pursued this internship 
with the aim of deepening understanding of advanced models, thereby enhancing 
my future research and analysis in the realm of multimodal cellular data.

 Over the past few months, I have embarked on an initial exploration of large models, delving into the development of LLMs across various domains.
 I’ve scrutinized the characteristics of raw data for our projects, begun engaging in prompt engineering, and developed effective prompts.
 I’ve utilized OpenAI’s interfaces to develop assistant and completion strategies that to generate corpus in bulk.
 Leveraging the LLaMA-Factory framework and the high-performing Chinese base model, Qwen, we’ve fine-tuned our first version of the model, which is still undergoing continuous research and improvement.
 In 2024, we aim to explore and research the application of agent models in specific domains.
 I will gradually share my accumulated knowledge in this vast field here. Please stay tuned! And if you are interested in LLMs, please feel free to contact me at [email protected]

  1. Two Modes of Calling the OpenAI API Interface: completion & assistant.
  2. Developing an agent that utilizes InternLM utilizing the MMSeg and MMDet libraries to analyze images.
  3. Research on LLM Agent.