西南科技大学SWUST

Welcome to join our family to learn together and promote the advancement of machine learning in China! Now, let's start {

1. If you don't have a high-performance GPU, I recommend that you rent one. There are cheap GPUs available for students on the Internet, even some are free.

2. Some students may not have a foundation in machine learning, but not need to be nervous. If you just want to know how to use large models, it's still easy.

3. follow the step, and you will have a basic understanding of the use of large models.
   

   "Tool": ||-python-||-pytorch-||-cuda-||-anaconda(miniconda)-||-pycharm(vscode)-||. I think it is easy for you, and there are many course on bilibili.
   "usage":
   >>first --- download "Transformer library","Tokenizer","Pretrained Model",and you can use Tsinghua-source(清华源) and hf-mirror to download them.
   >>second --- |"import"| -> |"Tokennizer"| -> |load "Pretrained-model"| -> |input your sentence or image| -> |"model.generate"|
   >>>>example:
   >>>>||----from transformers import GPT2Tokenizer, GPT2Model--||
   >>>>||----tokenizer = GPT2Tokenizer.from_pretrained('gpt2')------||
   >>>>||----model = GPT2Model.from_pretrained('gpt2')--------------||
   >>>>||----text = "Replace me by any text you'd like."------------------||
   >>>>||----encoded_input = tokenizer(text, return_tensors='pt')---||
   >>>>||----output = model.generate(encoded_input)------------------||
   "customized": It's not a easy job. But I can give a tips that you can start with Lora. Lora as PEFT is friendly for students. And there are other ways to fine-tune the model like prefix-tuning,P-tuning,RLHF,etc. Also you can try Data mounting.
   

} Nothing is difficult to the man who will try!