Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

LM FT #410

Open
junxnone opened this issue Jun 5, 2023 · 0 comments
Open

LM FT #410

junxnone opened this issue Jun 5, 2023 · 0 comments

Comments

@junxnone
Copy link
Owner

junxnone commented Jun 5, 2023

大模型微调



- Full Fine-Tuning
- Adapter-Tuning
  - LoRAs
    - Initialization
      - PiSSA
      - OLoRA
      - EVA
      - LoftQ
      - rsLoRA
      - DoRA
    - LoHa
    - LoKr
    - AdaLoRA
    - X-LoRA
  - OFT
  - BOFT
  - Llama-Adapter
  - HRA
  - Bone 
- Prompting
  - 提示调整(prompt tuning)
  - 前缀调整(prefix tuning)
  - P 调整(P-tuning)
  - 多任务提示调整(multitask prompt tuning)
  - Context-Aware Prompt Tuning (CPT)
- Dreambooth

Adapter-Tuning

在预训练模型的特定层或位置插入一些小型的适配器模块(Adapter)。在微调过程中,主要对这些适配器模块的参数进行调整,而保持预训练模型的大部分参数不变。通过这种方式,可以利用预训练模型的强大表示能力,同时以较小的计算成本和数据量实现对特定任务的适配和优化,使得模型在新任务上的性能得到提升,并且在一定程度上缓解过拟合现象,提高模型的泛化能力。

Prompting

  • 硬提示(hard prompts)是带有离散输入标记的手工制作的文本提示;缺点是创建一个好的提示需要很多努力。
  • 软提示(soft prompts)是与输入嵌入连接的可学习张量,可以针对数据集进行优化;缺点是它们不是人类可读的,因为没有将这些 “虚拟标记” 与真实单词的嵌入相匹配。

Reference

@junxnone junxnone changed the title Hot HugeModel Tuning Hot LLM Tuning Jun 13, 2023
@junxnone junxnone changed the title Hot LLM Tuning Hot LM Tuning Jun 13, 2023
@junxnone junxnone changed the title Hot LM Tuning LM FT Jan 24, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant