怎么查网站的浏览量,网站建设做网站需要多少钱,营销方式有哪几种,国际局势最新消息今天GPT-4 Vision 系列:
翻译: GPT-4 with Vision 升级 Streamlit 应用程序的 7 种方式一翻译: GPT-4 with Vision 升级 Streamlit 应用程序的 7 种方式二翻译: GPT-4 Vision静态图表转换为动态数据可视化 升级Streamlit 三翻译: GPT-4 Vision从图像转换为完全可编辑的表格 升级St…GPT-4 Vision 系列:
翻译: GPT-4 with Vision 升级 Streamlit 应用程序的 7 种方式一翻译: GPT-4 with Vision 升级 Streamlit 应用程序的 7 种方式二翻译: GPT-4 Vision静态图表转换为动态数据可视化 升级Streamlit 三翻译: GPT-4 Vision从图像转换为完全可编辑的表格 升级Streamlit四翻译: GPT-4 Vision通过量身定制的推荐来增强应用的用户体验 升级Streamlit五
1. 征服LLM幻觉hallucinations
毫无疑问GPT-4 比其前辈有了重大改进。
像所有 LLM 一样它可以产生误导性或虚构的输出称为幻觉hallucinations。这可能是由于训练数据中的偏差、提示不明确或 GPT-4 可能不包含最新数据这一事实。
这时检索增强生成 Retrieval Augmented Generation或RAG就派上用场了。RAG 是一种通过整合外部数据来改进聊天机器人的技术确保更相关和最新的响应。
例如GPT-4 不知道 Streamlit 的新彩色标题因为它们在训练时不可用。
我们将首先粘贴新的 st.header() 文档的屏幕截图其中包括用于着色标题的新 API 参数 包括以下提示prompt Build a Streamlit app featuring various st.header() in different colors, using the new divider argument.Include a brief humorous data science pun for each header.Add a corresponding emoji to each header. 使用新的divider参数构建一个具有不同颜色的各种 st.header() 的 Streamlit 应用程序。为每个标题添加一个简短的幽默数据科学双关语。为每个标题添加相应的表情符号。
让我们看看结果
Vision 在无缝显示彩虹标题方面做得很好。
有趣的事实与通过 Code Interpreter 上传 PDF 相比使用 Vision 上传文档屏幕截图会带来更好的聊天对话。
2. GPT-4 实现
根据你提供的截图显示Streamlit中的st.header()函数现在支持divider参数可以使用blue、“green”、“orange”、“red”、“violet”、gray/grey或rainbow等颜色数值。
这是一个更新后的Streamlit应用程序版本其中包含了不同颜色分隔线的st.header()调用以及每个标题对应的幽默数据科学双关语和相应的表情符号。
import streamlit as st# Set up the page layout
st.set_page_config(layoutwide)# Header with blue divider
st.header(Data scientists do it better with Python :snake:, dividerblue)
st.caption(Heres a plot twist for you!)# Header with green divider
st.header(Why was the computer cold at the data party? It left its Windows open :snowflake:, dividergreen)
st.caption(Dont freeze up on your data!)# Header with orange divider
st.header(Why dont data scientists get sunburned? They use plenty of layers for protection! :sun_with_face:, dividerorange)
st.caption(Dont forget to add layers to your neural networks too!)# Header with red divider
st.header(Data scientists are the real random forest rangers :deciduous_tree:, dividerred)
st.caption(Keeping the overfitting beasts at bay!)# Header with violet divider
st.header(A data scientists favorite mode of transport? The dataset! :train:, dividerviolet)
st.caption(Choo-choo-choose your parameters wisely!)# Header with grey divider
st.header(Breaking up with overfitting models is tough. They always predict you’ll come back! :broken_heart:, dividergrey)
st.caption(Its time to regularize your relationships!)# Header with rainbow divider
st.header(If youre a SQL database, I’d join you with no conditions :heart_eyes:, dividerrainbow)
st.caption(The best relationships are the ones without keys!)# Note: The divider argument is used as per the screenshot provided to set the color of the divider for each header.
确保你正在运行支持st.header()函数中divider参数的Streamlit版本就像你的截图一样。要查看这些带有分隔线和标题的标题渲染请将脚本保存在Python文件中例如headers_app.py并使用Streamlit运行它。
streamlit run Vision/headers_app.py 这将生成一个Streamlit页面其中包含多个标题每个标题都配有一个幽默的数据科学双关语和一个表情符号用指定颜色的分隔线分隔。
代码
https://github.com/zgpeace/pets-name-langchain/tree/develop
参考
https://blog.streamlit.io/7-ways-gpt-4-vision-can-uplevel-your-streamlit-apps/