【星海出品】LangChain-Installation

Environment: python3.10

Install langchain

Mannagment
docs

Record In 2024/05/30
support pip or conda

#
# this package is the essential
#
pip install langchain
#
# Third party expansion package
#
pip3 install langchain-community  -i http://aaa.aaa.aaa.aaa/repository/pypi/simple --trusted-host aaa.aaa.aaa.aaa
#
Collecting langchain-community
  Downloading http://aaa.aaa.aaa.aaa:xxxx/repository/pypi/packages/langchain-community/0.2.1/**langchain_community-0.2.1-py3-none-any.whl** (2.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.1/2.1 MB 76.0 MB/s eta 0:00:0
Collecting dataclasses-json<0.7,>=0.5.7 (from langchain-community)
  Downloading http://aaa.aaa.aaa.aaa:xxxx/repository/pypi/packages/dataclasses-json/0.6.6/dataclasses_json-0.6.6-py3-none-any.whl (28 kB)
Collecting marshmallow<4.0.0,>=3.18.0 (from dataclasses-json<0.7,>=0.5.7->langchain-community)
  Downloading http://aaa.aaa.aaa.aaa:xxxx/repository/pypi/packages/marshmallow/3.21.2/marshmallow-3.21.2-py3-none-any.whl (49 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 49.3/49.3 kB 11.3 MB/s eta 0:00:00
Collecting typing-inspect<1,>=0.4.0 (from dataclasses-json<0.7,>=0.5.7->langchain-community)
Collecting mypy-extensions>=0.3.0 (from typing-inspect<1,>=0.4.0->dataclasses-json<0.7,>=0.5.7->langchain-community)
  Downloading http://aaa.aaa.aaa.aaa:xxxx/repository/pypi/packages/mypy-extensions/1.0.0/mypy_extensions-1.0.0-py3-none-any.whl (4.7 kB)
Installing collected packages: mypy-extensions, marshmallow, typing-inspect, dataclasses-json, langchain-community

pip3 install langchain-core
# [prompt]
# langchain package contain langchain-core
#

#
# openai is import
#
pip install langchain-openai

#
# update to latest
#
pip install --upgrade --quiet  langchain langchain-community langchainhub langchain-openai langchain-chroma bs4

pip install langchain-community
pip install langchain-experimental
pip install langgraph

conda

conda install langchain -c conda-forge

Source Install

Dir: PATH/TO/REPO/langchain/libs/langchain

pip install -e .

如果想要访问还是要有代理的

import os
os.environ['http_proxy'] = 'http://127.0.0.1:xxxx'
os.environ['https_proxy'] = 'https://127.0.0.1:xxxx'

from langchain.llms import OpenAI
# llm = OpenAI(openai_api_key="...")
llm = OpenAI()

Demo1

pip install -qU langchain-openai

import os
import getpass
os.environ["OPENAI_API_KEY"] = getpass.getpass()

from langchain_openai import ChatOpenAI

model = ChatOpenAI(model="gpt-4")

from typing import Optional

from langchain_core.pydantic_v1 import BaseModel, Field


class Joke(BaseModel):
    """Joke to tell user."""

    setup: str = Field(description="The setup of the joke")
    punchline: str = Field(description="The punchline to the joke")
    rating: Optional[int] = Field(description="How funny the joke is, from 1 to 10")


structured_llm = llm.with_structured_output(Joke)

structured_llm.invoke("Tell me a joke about cats")

Joke(setup=‘Why was the cat sitting on the computer?’, punchline=‘To keep an eye on the mouse!’, rating=None)

Demo2

pip install openai chromadb tiktoken

Win+R [control system]
set environment

GNN Frameworks（图神经网络框架）

GNNAdvisor

GNNAdvisor使用Pytorch作为前端，在底层是用CUDA写的，并用pytorch wrapper集成到pytorch里面。
它可以被看作是具有kernel优化、运行时支持的operator

embedding 并行

GNN的每个节点都有维度很大的embedding
维度可以调整

Pytorch、TensorFlow，内置了许多NN operator，针对欧几里得数据做了很多优化，但是对于图这样的非欧几里得数据，面临一些挑战：

• torch-scatter在处理大规模稀疏图和高维节点embedding时性能不佳，问题在于其kernel采用了高开销的原子操作来执行节点embedding传播，这种方法借鉴了图处理系统的设计原则，但在面临大规模稀疏图与高维节点embedding时性能不佳；
• DGL采用现成的SPMM（比如cuSparse库的csrmm2）来进行sum-reduced aggregation，同时利用自己的cuda kernel处理带有边属性的聚合。

欧几里得数据
three_3d_array = np.random.randint((3, 3, 3))
three_3d_array = [[[8 4 3]
[7 6 1]
[8 3 2]]
[[2 5 8]
[2 3 6]
[0 7 4]]
[[8 6 6]
[3 6 2]
[9 2 0]]]
使用numpy进行数据的调度

# 访问位于索引(1, 1, 1)的元素  
element = random_3d_array[1, 1, 1]
print(element)
# 访问第一层的所有元素  
first_layer = random_3d_array[0, :, :]
print("layer:",first_layer)
# 访问第一列的所有元素  
first_column = random_3d_array[:,:, 0]
print("column:",first_column)