Can You Tell Me About DSPy: How DSPy Automates Prompt Engineering Modularly
ฝัง
- เผยแพร่เมื่อ 29 มิ.ย. 2024
- Demonstrate-Search-Predict (DSP), with 'Py'thon
Developed by Stanford NLP Group
DSPy provides general purpose modules that replace string-based
prompting. DSPy also provides optimizers which can optimize these
modules with help of following
What are the various modules DSPy has?
Lang Models, Signatures, Modules, Data, Metrics, Optimizers, Assertions
Can you talk about use cases of DSPy?
QA, Classification, Summarisatio, RAGs / Multi-Hop rags,
Reasoning
But can you tell me Why DSPy?
- Streamline the process of prompt-engineering with input and objective
- Create Modules of LMs that can work to achieve objective
- Then train the modules by using Datasets built by hand, and with LLMs
- Evalute the modules with the Metrics that is custom coded in DSPy
- Including Fine-tuning smaller models like T5 and Bert
Chapter Navigation:
0:00 Intro
0:50 Video Overview
2:00 DSPy Way
3:10 What DSPy does
5:25 Modules in DSPy
6:10 Tasks Done in DSPy
6:50 Installing DSPy-AI
7:50 Use Virtual Environment
9:10 Minimal Example of DSPY
10:10 Creating & Prompting LM object
12:50 Creating Predict Object
14:20 TypeError in Predict
15:00 Prediction Object
16:15 Reviewing LM Calls History
16:40 Reviewing Prompt Signature
17:35 Programming Prompts
18:00 How Prompts are Programmed
18:45 What is Next
19:25 Why DSPy?
20:30 Outro
The data and the code is located at
github.com/insightbuilder/pyt...
I believe you will like this video, and subscribe to the channel. Further uploads related to Big Data, Large Language models and Artificial Intelligence will be shared to your TH-cam Dashboard Directly.
The supporting playlists are
The bard Project
• Google Bard LLM : New ...
Practical Projects Playlist
• The Future is Here: La...
Huggingface Playlist
• Mastering NLP with Hug...
Python Data Engineering Playlist
• Learn to Data Engineer...
Python Ecosystem of Libraries
• Mastering the Python E...
ChatGPT and AI Playlist
• Learn about AI Languag...
AWS and Python AWS Wrangler
• Building a Powerful Da...
Exploring the Realm of Generative AI: Hardware and Software Discussions
• Exploring the Realm of...
PS: Got a question or have a feedback on my content. Get in touch
By leaving a Comment in the video
topmate.io/insightbuilder?SocialProfile
@mail insighthacker21@gmail.com
@twitter Handle is @KQrios
@medium / about
@github github.com/Kamalabot - วิทยาศาสตร์และเทคโนโลยี
Py is for pytorch. Great series
Excellent video! You made a very complex subject easier to be digested ✌🏼✌🏼
Nice. Explaining how the implementation is done in dspy was helpful. Keep up the good work my friend
Excited for this series. Thanks for these videos!
Yep, it's exciting library indeed. DSPy optimizers are a step towards making prompt engineering and collecting data for fine-tuning streamlined.
Learning about individual parts of DSPy is key. Happy learning
Very well explained, keep up the good work