Skip to content

Commit f7fc0fe

Browse files
Java-EdgeJava-Edge
committed
docs:专栏更新
1 parent 97e29f4 commit f7fc0fe

13 files changed

+2392
-323
lines changed

‎.vscode/.server-controller-port.log‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
{
22
"port": 9145,
3-
"time": 1728450843494,
3+
"time": 1729093610425,
44
"version": "0.0.3"
55
}

‎docs/.vuepress/config.js‎

Lines changed: 39 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -516,9 +516,17 @@ module.exports = {
516516
{
517517
text: 'Redis',
518518
items: [{
519-
text: 'Redis数据结构的最佳实践',
519+
text: '基础',
520520
link: '/md/redis/00-数据结构的最佳实践.md'
521-
}]
521+
},
522+
{
523+
text: '源码',
524+
link: '/md/redis/00-数据结构的最佳实践.md'
525+
},
526+
{
527+
text: '业务',
528+
link: '/md/redis/00-数据结构的最佳实践.md'
529+
},]
522530
},
523531

524532
{
@@ -1110,6 +1118,7 @@ module.exports = {
11101118
collapsable: false,
11111119
sidebarDepth: 0,
11121120
children: [
1121+
"04-RPC框架在网络通信的网络IO模型选型",
11131122
"熔断限流",
11141123
"11-RPC的负载均衡",
11151124
]
@@ -1176,7 +1185,7 @@ module.exports = {
11761185
children: [
11771186
"为啥要学习数据分析？",
11781187
"correct-data-analysis-learning-methods",
1179-
"02-数据挖掘的学习路径",
1188+
"learning-path-data-mining",
11801189
"企业如何利用数据打造精准用户画像？",
11811190
"如何自动化采集数据",
11821191
"how-to-use-octoparse-for-data-scraping",
@@ -1194,6 +1203,7 @@ module.exports = {
11941203
"03-ReentrantLock与AQS.md",
11951204
"04-线程池以及生产环境使用.md",
11961205
"05-京东并行框架asyncTool如何针对高并发场景进行优化？.md",
1206+
"java21-virtual-threads-where-did-my-lock-go",
11971207
]
11981208
},
11991209
{
@@ -1562,6 +1572,7 @@ module.exports = {
15621572
collapsable: false,
15631573
sidebarDepth: 0,
15641574
children: [
1575+
"01-Netty源码面试实战+原理(一)-鸿蒙篇",
15651576
"netty-basic-components",
15661577
"ChannelPipeline接口",
15671578
"(06-1)-ChannelHandler 家族",
@@ -1664,20 +1675,34 @@ module.exports = {
16641675
}, ],
16651676

16661677
"/md/redis/": [{
1667-
title: "Redis",
1678+
title: "基础",
16681679
collapsable: false,
16691680
sidebarDepth: 0,
16701681
children: [
1671-
"00-数据结构的最佳实践",
16721682
"01-Redis和ZK分布式锁优缺点对比以及生产环境使用建议",
1673-
"02-Redisson可重入锁加锁源码分析",
1674-
"03-Redisson公平锁加锁源码分析",
1675-
"04-Redisson读写锁加锁机制分析",
16761683
"05-缓存读写策略模式详解",
16771684
"06-如何快速定位 Redis 热 key",
16781685
"12-Redis 闭源？",
16791686
]
1680-
}],
1687+
},
1688+
{
1689+
title: "源码",
1690+
collapsable: false,
1691+
sidebarDepth: 0,
1692+
children: [
1693+
"02-Redisson可重入锁加锁源码分析",
1694+
"03-Redisson公平锁加锁源码分析",
1695+
"04-Redisson读写锁加锁机制分析",
1696+
]
1697+
},
1698+
{
1699+
title: "业务",
1700+
collapsable: false,
1701+
sidebarDepth: 0,
1702+
children: [
1703+
"00-数据结构的最佳实践",
1704+
]
1705+
},],
16811706
"/md/es/": [{
16821707
title: "ElasticSearch",
16831708
collapsable: false,
@@ -2076,6 +2101,7 @@ module.exports = {
20762101
"11-lcel-memory-addition-method",
20772102
"12-lcel-agent-core-components",
20782103
"13-best-development-practices",
2104+
"local-large-model-deployment",
20792105
]
20802106
},
20812107

@@ -2087,6 +2113,9 @@ module.exports = {
20872113
"01-three-minute-fastapi-ai-agent-setup",
20882114
"02-Agent应用对话情感优化",
20892115
"03-use-tts-to-make-your-ai-agent-speak",
2116+
"langserve-revolutionizes-llm-app-deployment",
2117+
"customizing-a-tool-for-your-ai-agent",
2118+
"Complex-SQL-Joins-with-LangGraph-and-Waii",
20902119
"AI Agent应用出路到底在哪？",
20912120
]
20922121
},
@@ -2098,6 +2127,7 @@ module.exports = {
20982127
children: [
20992128
"00-introduce-to-LangGraph",
21002129
"langgraph-studio",
2130+
"multi_agent",
21012131
"methods-adapting-large-language-models",
21022132
"to-fine-tune-or-not-to-fine-tune-llm",
21032133
"effective-datasets-fine-tuning",

‎docs/md/AI/13-best-development-practices.md‎

Lines changed: 1 addition & 202 deletions
Original file line numberDiff line numberDiff line change
@@ -27,205 +27,4 @@
2727

2828
优点：数据私有、更灵活、成本低
2929

30-
缺点：算力设施、技术支撑
31-
32-
## 3 使用 Ollama 在本地部署大模型
33-
34-
### 3.1 下载并运行应用程序
35-
36-
37-
38-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/a872adde1e96e5dbd3ddb0e910f48088.png)
39-
40-
41-
42-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/de0bfb92df17722ebdbb5c0696fd7666.png)
43-
44-
45-
46-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/daa95f47315ba60e6790d27661f85021.png)
47-
48-
49-
50-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/85b62d04db4c06665b1fff64de5bec87.png)
51-
52-
### 3.2 从命令行中选取模型(ollama pull llam2)
53-
54-
[官网支持的模型](https://ollama.com/library?sort=newest)：
55-
56-
![](/Users/javaedge/Downloads/IDEAProjects/java-edge-master/assets/image-20240621135627185.png)
57-
58-
挑选一个比较小的试玩下：
59-
60-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/46b83f44f00fb3965c35e700cb45eb85.png)
61-
62-
### 3.3 运行
63-
64-
[浏览器](localhost:11434)：
65-
66-
![](/Users/javaedge/Downloads/IDEAProjects/java-edge-master/assets/image-20240621141710055.png)
67-
68-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/29fa4e05840db498501e59e03db1e63f.png)
69-
70-
## 4 本地大模型调用
71-
72-
既然部署本地完成了，来看看如何调用呢？
73-
74-
```python
75-
from langchain_community.llms import Ollama
76-
77-
llm = Ollama(model="qwen2:0.5b")
78-
llm.invoke(input="你是谁？")
79-
```
80-
81-
82-
83-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/af07e34926600fdd9946e2905c05bb7a.png)
84-
85-
### 使用流式
86-
87-
```python
88-
#使用流式
89-
from langchain.callbacks.manager import CallbackManager
90-
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
91-
92-
llm = Ollama(
93-
model="qwen2:0.5b", callback_manager=CallbackManager([StreamingStdOutCallbackHandler()])
94-
)
95-
llm.invoke(input="第一个登上月球的人是谁?")
96-
```
97-
98-
99-
100-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/57bcab3fd266daac316d119b20199b37.png)
101-
102-
## 5 模型评估
103-
104-
### 5.1 远程大模型
105-
106-
```python
107-
from langchain_openai import ChatOpenAI
108-
from langchain.evaluation import load_evaluator
109-
llm = ChatOpenAI(
110-
api_key=os.getenv("DASHSCOPE_API_KEY"),
111-
base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
112-
model="qwen-plus"
113-
)
114-
115-
evaluator = load_evaluator("criteria", llm=llm, criteria="conciseness")
116-
eval_result = evaluator.evaluate_strings(
117-
prediction="four.",
118-
input="What's 2+2?",
119-
)
120-
print(eval_result)
121-
```
122-
123-
124-
125-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/bb4d72b250043b2ee5bd0ae82541e655.png)
126-
127-
如果不简洁的回答：
128-
129-
```python
130-
#inpt 输入的评测问题
131-
#prediction 预测的答案
132-
# 返回值 Y/N 是否符合
133-
# 返回值score 1-0分数，1为完全，0为不完全
134-
eval_result = evaluator.evaluate_strings(
135-
prediction="What's 2+2? That's an elementary question. The answer you're looking for is that two and two is four.",
136-
input="What's 2+2?",
137-
)
138-
print(eval_result)
139-
```
140-
141-
142-
143-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/61c1b940051d6c7b5849cf6211fceefb.png)
144-
145-
### 5.2 本地大模型
146-
147-
```python
148-
from langchain_community.chat_models import ChatOllama
149-
llm = ChatOllama(model="qwen2:0.5b")
150-
evaluator = load_evaluator("criteria", llm=llm, criteria="conciseness")
151-
```
152-
153-
```python
154-
#inpt 输入的评测问题
155-
#prediction 预测的答案
156-
# 返回值 Y或者N是否符合
157-
# 返回值score 1-0分数，1为完全，0为不完全
158-
eval_result = evaluator.evaluate_strings(
159-
prediction="What's 2+2? That's an elementary question. The answer you're looking for is that two and two is four.",
160-
input="What's 2+2?",
161-
)
162-
print(eval_result)
163-
```
164-
165-
166-
167-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/ea116b14383b6db7194d7658810767fd.png)
168-
169-
### 5.3 内置评估标准
170-
171-
```python
172-
# 内置的一些评估标准
173-
from langchain.evaluation import Criteria
174-
175-
list(Criteria)
176-
```
177-
178-
179-
180-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/f71d5563c6a00a18f2951bb3a4e2f9cd.png)
181-
182-
183-
184-
185-
```python
186-
llm = ChatOllama(model="qwen2:0.5b")
187-
#使用enum格式加载标准
188-
from langchain.evaluation import EvaluatorType
189-
#自定义评估标准
190-
custom_criterion = {
191-
"幽默性": "输出的内容是否足够幽默或者包含幽默元素",
192-
}
193-
eval_chain = load_evaluator(
194-
EvaluatorType.CRITERIA,
195-
llm=llm,
196-
criteria=custom_criterion,
197-
)
198-
query = "给我讲一个笑话"
199-
prediction = "有一天，小明去买菜，结果买了一堆菜回家，结果发现自己忘了带钱。"
200-
eval_result = eval_chain.evaluate_strings(prediction=prediction, input=query)
201-
print(eval_result)
202-
```
203-
204-
205-
206-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/b626bd419b59ded036872353dbd91d41.png)
207-
208-
### 模型比较
209-
210-
```python
211-
from langchain.model_laboratory import ModelLaboratory
212-
from langchain.prompts import PromptTemplate
213-
from langchain_openai import OpenAI
214-
from langchain_community.llms.chatglm import ChatGLM
215-
from langchain_community.chat_models import ChatOllama
216-
217-
#比较openai、ChatGLM、ChatOllama三个模型的效果
218-
llms = [
219-
# OpenAI(temperature=0),
220-
ChatOllama(model="qwen2:0.5b"),
221-
]
222-
```
223-
224-
```python
225-
model_lab = ModelLaboratory.from_llms(llms)
226-
model_lab.compare("齐天大圣的师傅是谁？")
227-
```
228-
229-
230-
231-
![](https://my-img.javaedge.com.cn/javaedge-blog/2024/06/8c693bac93ab5309068b4a724dd9eac1.png)
30+
缺点：算力设施、技术支撑

0 commit comments

Comments
 (0)