91精品在线影院,中文字幕久久久,久久久久一区二区三区四区

��球数据资源

paulwong — Thu, 24 Apr 2025 06:56:00 GMT

��球基础数据

https://www.nami.com/details/4nw10i0tela68lq#interface

��球�l�计数据

https://www.nami.com/details/7xwk3iqtv3s9rk6#interface

��球�l�计数据

https://www.nami.com/details/7xwk3iqtv3s9rk6#interface

��球高阶数据

https://www.nami.com/details/g5wvvikteeixwzd#interface

指数数据

https://www.nami.com/details/o6w9kipt4yi78k3#interface

��球资料库数�?/div>

https://www.nami.com/details/7j8gxi0to7inrql#interface

Marz火星数据�Q�体�Ԍ��

https://www.kancloud.cn/marz/marz-sport/3098904

paulwong 2025-04-24 14:56 发表评论

大模型训�l�的几个阶段

paulwong — Tue, 18 Mar 2025 05:14:00 GMT

大模型开发出来后, 一般要�l�过以下几个阶段的训�l?

预训�l?Pre-Trained)

单纯提供文本: {"text":"..."}

训练模型��q��一个文字开�? 预测后面的文�? 直到�l�束.

�q�种模型只会做完成文本的��d��

监督微调(Supervised Fine Turning)

��Z��使模型能完成�Ҏ��指��o完成回答, 而不是随机生成回�{?/div>

提供的文�? {"instruction":"...", "output":"..."}

高效参数微调(Parameter Efficient Fine Turning)

只调整部分参�? 具体实现�Ҏ��有LoRA

参�?

https://github.com/huggingface/smol-course/blob/main/1_instruction_tuning/notebooks/sft_finetuning_example.ipynb

paulwong 2025-03-18 13:14 发表评论

paulwong — Wed, 12 Mar 2025 02:08:00 GMT

大模型微调后的评估指标是衡量模型性能的关键，通常�Ҏ��d��c�d��和具体需求选择不同的评估指标。以下是一些常见的评估指标及其适用场景�Q?/p>

1. 分类��d��

准确率（Accuracy�Q?/span>�Q�预��正��的��h��占��L��本的比例�?
- 适用场景�Q�类别分布均衡的��d��?/li>
�_��率（Precision�Q?/span>�Q�预��ؓ正类的样本中�Q�实际�ؓ正类的比例�?
- 适用场景�Q�关注减��假��x��（False Positive�Q�的��d��?/li>
召回率（Recall�Q?/span>�Q�实际�ؓ正类的样本中�Q�预��ؓ正类的比例�?
- 适用场景�Q�关注减��假阴性（False Negative�Q�的��d��?/li>
F1分数�Q�F1 Score�Q?/span>�Q�精��率和召回率的调和��^均倹{�?
- 适用场景�Q�类别不�q��或需要��^衡精��率和召回率的�Q务�?/li>
ROC-AUC�Q�ROC曲线下的面积�Q�衡量模型区分正负类的能力�?
- 适用场景�Q�二分类��d��Q�尤其是�c�d��不��^衡的情况�?/li>

2. 回归��d��

均方误差�Q�MSE, Mean Squared Error�Q?/span>�Q�预��g��真实��g��差的�q�x��的��^均倹{�?
- 适用场景�Q�对误差较大的样本惩�|�更重的��d��?/li>
均方根误差（RMSE, Root Mean Squared Error�Q?/span>�Q�MSE的��^�Ҏ��?
- 适用场景�Q�与MSE�c�M��Q�但更接�q�原始数据尺度�?/li>
�q�_��l�对误差�Q�MAE, Mean Absolute Error�Q?/span>�Q�预��g��真实��g��差的�l�对值的�q�_��倹{�?
- 适用场景�Q�对异常��g��敏感的�Q务�?/li>
R²�Q�决定系敎ͼ��Q�模型解释目标变量方差的比例�?
- 适用场景�Q�评估模型拟合优度�?/li>

3. 生成��d��

BLEU�Q�Bilingual Evaluation Understudy�Q?/span>�Q�衡量生成文本与参考文本的n-gram重叠�E�度�?
- 适用场景�Q�机器翻译、文本生成�Q务�?/li>
ROUGE�Q�Recall-Oriented Understudy for Gisting Evaluation�Q?/span>�Q�衡量生成文本与参考文本的重叠�E�度�Q�侧重于召回率�?
- 适用场景�Q�文本摘要、生成�Q务�?/li>
METEOR�Q�综合考虑�_��率、召回率和词序的评估指标�?
- 适用场景�Q�机器翻译、文本生成�Q务�?/li>
Perplexity�Q�困惑度�Q?/span>�Q�衡量模型预��概率分布的不确定性�?
- 适用场景�Q�语�a�模型评估�?/li>

4. 多标�{��Q�?/span>

Hamming Loss�Q�预��错误的标签比例�?
- 适用场景�Q�多标签分类��d��?/li>
Jaccard Similarity�Q�预��标�{�与真实标签的交集与�q��之比�?
- 适用场景�Q�多标签分类��d��?/li>

5. 排序��d��

NDCG�Q�Normalized Discounted Cumulative Gain�Q?/span>�Q�衡量排序结果的相关性�?
- 适用场景�Q�推荐系�l�、信息检索�?/li>
MAP�Q�Mean Average Precision�Q?/span>�Q��^均精��率的均倹{�?
- 适用场景�Q�信息检索、推荐系�l��?/li>

6. 其他指标

训练旉��Q�模型微调所需的时间�?/li>
推理速度�Q�模型生成结果的速度�?/li>
资源消�?/span>�Q�模型运行所需的计��资源（如GPU内存、CPU使用率）�?/li>
鲁棒�?/span>�Q�模型对噪声、异常值或�Ҏ��h��的抵抗能力�?/li>

7. 领域特定指标

��d��领域�Q�敏感性（Sensitivity�Q�、特异性（Specificity�Q�、AUC-ROC�?/li>
金融领域�Q�收益曲�Uѝ��夏普比率（Sharpe Ratio�Q��?/li>
计算��?/span>�Q�mAP�Q�mean Average Precision�Q�、IoU�Q�Intersection over Union�Q��?/li>

8. 人类评估

人工评分�Q�通过人工评估生成�l�果的质量（如流畅性、相��x��、准��性）�?/li>
用户满意�?/span>�Q�通过用户反馈评估模型的实际效果�?/li>

9. 模型�Ҏ��

基线�Ҏ��Q�与未微调的模型或基�U�模型进行性能�Ҏ��?/li>
消融实验�Q�评估微调过�E�中不同�l��g�Q�如数据、超参数�Q�对性能的媄响�?/li>

10. �l�合评估

多指标综�?/span>�Q�根据�Q务需求，�l�合多个指标�q�行�l�合评估�?/li>
��d��特定指标�Q�针对特定�Q务设计自定义指标�?/li>

在实际应用中�Q�选择合适的评估指标需要结合�Q务目标、数据特点和业务需求，同时注意避免单一指标的局限性�?/p>

paulwong 2025-03-12 10:08 发表评论

LLM全栈框架完整分类清单�Q�预训练+微调+工具链）

paulwong — Mon, 10 Mar 2025 03:29:00 GMT

https://blog.csdn.net/ViniJack/article/details/145789900

paulwong 2025-03-10 11:29 发表评论

�ȝ��问诊�pȝ��资源

paulwong — Sat, 08 Mar 2025 12:52:00 GMT

计算机毕业设计Python+Neo4j知识图谱�ȝ��问答�pȝ�� 大模�?br />

https://baijiahao.baidu.com/s?id=1815574648931972744&wfr=spider&for=pc

QABasedOnMedicaKnowledgeGraph

https://github.com/liuhuanyong/QASystemOnMedicalKG/blob/master/README.md

非结构文字抽取实体与关系的大模型

底��, 癑ַ� https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/tree/main

底��, llama2 https://huggingface.co/unsloth/llama-2-13b

微调->癑ַ� https://huggingface.co/zjunlp/baichuan2-13b-iepile-lora

微调->llama2 https://huggingface.co/zjunlp/llama2-13b-iepile-lora

SiameseUniNLU通用自然语言理解模型

https://www.modelscope.cn/models/iic/nlp_structbert_siamese-uninlu_chinese-base/summary

数据�?/div>

https://huggingface.co/datasets/zjunlp/iepile

各种已经训练好的模型

https://www.modelscope.cn/models?name=zpeng1989&page=1

paulwong 2025-03-08 20:52 发表评论

使用nlp提取非结构化数据中的信息

paulwong — Sat, 08 Mar 2025 03:45:00 GMT

@import url(http://www.aygfsteel.com/CuteSoft_Client/CuteEditor/Load.ashx?type=style&file=SyntaxHighlighter.css);@import url(/css/cuteeditor.css); @import url(http://www.aygfsteel.com/CuteSoft_Client/CuteEditor/Load.ashx?type=style&file=SyntaxHighlighter.css);@import url(/css/cuteeditor.css); 如果要从�l�构化的数据中提取信�?用sql卛_��, 卌��提取的信息在select 的字�D�中.

如果要从非结构化的数据中, 如纯文本, 则要靠nlp, 要对文本理解�? 才能提取相应的信�?

https://www.w3cschool.cn/article/99991254.html

文本�l�构�?with SpaCy �ȝ��

https://zhuanlan.zhihu.com/p/556163162

https://zhuanlan.zhihu.com/p/557953165

https://zhuanlan.zhihu.com/p/563334531

https://zhuanlan.zhihu.com/p/573743734

使用openspg自动构徏�ȝ��知识图谱

https://blog.csdn.net/myboyliu2007/article/details/139654943

paulwong 2025-03-08 11:45 发表评论

AI案例资源

paulwong — Wed, 26 Feb 2025 08:01:00 GMT

@import url(http://www.aygfsteel.com/CuteSoft_Client/CuteEditor/Load.ashx?type=style&file=SyntaxHighlighter.css);@import url(/css/cuteeditor.css);

从实跉|��例介�l�大模型应用�l�验和思�?/div>

https://mp.weixin.qq.com/s/hcD0-z9Y4PsrILUgHdqGcQ

LLaMA Factory�Q�微调DeepSeek-R1-Distill-Qwen-7B模型实现新闻标题分类�?

https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory_deepseek_r1_distill_7b

deepseek r1微调模型应用落地案例�Q�医疗法律，PatientSeek�Q?/div>

https://www.bilibili.com/video/BV17zAVevEtw/?spm_id_from=333.788.recommend_more_video.0&vd_source=35b81999db00535703a287d5c98652b1

文本转语音的模型ChatTTS体验极佳�Q�真��丝滑和流畅，自定义也比较灉|��

https://www.bilibili.com/video/BV1oJ4m1u7B8/?spm_id_from=333.1387.upload.video_card.click&vd_source=35b81999db00535703a287d5c98652b1

�ȝ��NLP领域评测/比赛�Q�数据集�Q�论文和预训�l�模型资源汇怅R�?/div>

https://github.com/FreedomIntelligence/Medical_NLP

paulwong 2025-02-26 16:01 发表评论

不用再找了，�q�是大模型最全的面试题库

paulwong — Tue, 21 Jan 2025 23:42:00 GMT

https://blog.csdn.net/m0_59596990/article/details/135200833

paulwong 2025-01-22 07:42 发表评论

paulwong — Fri, 17 Jan 2025 07:52:00 GMT

@import url(http://www.aygfsteel.com/CuteSoft_Client/CuteEditor/Load.ashx?type=style&file=SyntaxHighlighter.css);@import url(/css/cuteeditor.css);

https://hyper.ai/cn/datasets

paulwong 2025-01-17 15:52 发表评论

vllm资源

paulwong — Fri, 17 Jan 2025 05:01:00 GMT

vllm是一个可以加载大模型, 推理, 量化模型, 以http api的方式暴露服务的框架.

https://docs.vllm.ai/en/latest/getting_started/examples/basic_with_model_default_sampling.html

paulwong 2025-01-17 13:01 发表评论

AI应用场景

paulwong — Fri, 17 Jan 2025 03:23:00 GMT

到底AI是虚的还是假�? 在企业中有没实际落地场景, 以下取实际应用场�?

生物公司

使用qwen2:7b训练�l�胞制备领域的数据集�Q�目标是
1.预测�l�胞收获�?nbsp;
2.��细胞存�zȝ��?存活/��M��)
3.预测工艺是否成功
4.可以提前预测�l�胞的质量是否达标，以便及时采取措施�q�行调整
5.�l�胞培养�q�程中出现大量细胞死亡的情况�Q�模型可以根据实时数据和历史�l�验�Q�分析可能是培养��温度失控、培��d��成分错误或受到污染等原因��D��的，�q�提供相应的排查��?/div>

文体旅游

��旅游�pȝ��:
提供目的��C��l��?br /> 旅行路线规划�?br /> 酒店预订和景
�Ҏ��荐等服务�?/div>

考试改卷

��Z��大模型，做一个判试卷的应用，能够判断主观题，比如阅读理解�Q�比如历�Ԍ��地理�Q�政治问�{�题�?br /> 判卷准确率不能低于�h工判卷准��率�?br /> 即一�ơ考试�Q�一个班50份试��P��判断�l�果错误不超�q?道题。判断效率高于或�{�于人工�?br />
取过往同学试卷题目, 作答内容, 得分作一波ocr出数�? 一个科�? 提取所有试卷内�? 最后就是一个科目一个模�? 提取的内�Ҏ��在文�? csv, json,
��Z��“bert-base-chinese”�q�个模型, �q�行微调��Z��用模型即�?
让大模型成�ؓ专业的判卯��师

考试

用扣子打一个智能体�Q�实��C��同学员对掌握的知识进行测试，�Ҏ��试�l�果�q�行打分和二�ơ出题测�?/div>

paulwong 2025-01-17 11:23 发表评论

搭徏llamafactory微调、评估、测试和量化环境

paulwong — Thu, 16 Jan 2025 08:54:00 GMT

0. 配置环境变量

HF_ENDPOINT=https://hf-mirror.com
HF_HOME=/root/autodl-tmp/paul/tools/huggingface

1. 本机安装python 3.10, �q�设�|��Y件源

pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
pip config set global.index-url https://mirrors.huaweicloud.com/repository/pypi/simple

2. 安装miniconda

https://juejin.cn/post/7078965942968909854

3. 新徏一个环�? �q�激�z?/div>

conda create -n quantization python=3.12

2. 本机安装pytorch2.5.1+cuda12.4

pip3 install torch torchvision torchaudio

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

3. clone llamafactory源码

git clone https://github.com/hiyouga/LLaMA-Factory

4. llamafactory本地安装依赖

pip install -e .

pip install -e .["vllm","gptq"]

5. 启动webui

llamafactory-cli webui

6. 在页面中填入相关参数�q�行操作

paulwong 2025-01-16 16:54 发表评论

安装docker版的Nvidia container toolkit

paulwong — Mon, 13 Jan 2025 06:20:00 GMT

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installation

paulwong 2025-01-13 14:20 发表评论

AI入门

paulwong — Sat, 19 Oct 2024 14:37:00 GMT

数据分析�Q�从一堆已知的数据中进行分�c�，�ȝ��得出�l�计数据�Q�如最�?��|��最��|��q�_��|��d��{��?br /> 只能对已知数据进行操作，无法预测出新的数据的特征�Q�于是就有了机器学习�?

机器学习�Q�给��Z��堆已知的�Q�有特征栏位的和�l�果栏位的数据，选定一个算法，如线性回归，逻辑回归�{�，其实��是一条公式，�q�行学习�Q�其实就是运行一堆函敎ͼ�比较�l�果�Q�得��律，也就是确定了公式中参数的倹{��当输入新的数据�Ӟ��p��预测出所需的结果，其实��是把输入数据代入公式，��出�l�果�?/div>

机器学习只能做比较简单的��d��Q�如预测下个月的销售数据，判断文字内容是正面还是反�?分类)�Q�对于复杂的��d��Q�如对话�Q�其实就是针对输入文字预��靠��q��输出文字(回答)�Q�于是就有了深度学习�?/div>

深度学习�Q�给��Z��堆数据，只需两个本栏位，如问题，�{�案�{�，选定一个算法，其实��是��经�|�络的类型，如卷�U�神�l�网�l?CNN)�Q��@环神�l�网�l?RNN)�Q�TRANSFORMER��经�|�络�{�，�q�行学习�Q�其实就是运行一堆函敎ͼ�比较�l�果�Q�得��律，也就是确定了公式中参数的倹{�?img src ="http://www.aygfsteel.com/paulwong/aggbug/451501.html" width = "1" height = "1" />

paulwong 2024-10-19 22:37 发表评论

微调llama3大模�?2) - 使用ollama搭徏chatbot

paulwong — Mon, 08 Jul 2024 11:48:00 GMT

上篇已经合�ƈ��Z��训练好的大模型，现在要搭��v一套CHATBOT�Q��得这套大模型能有一个WEBUI用�v来�?/div>

1.讄��环境变量�Q�ollama的模型保存�\径，/etc/profile

export OLLAMA_MODELS=/root/autodl-tmp/models/ollama

2.克隆ollama代码

curl -fsSL https://ollama.com/install.sh | sh

3.启动ollama

ollama serve

4.建立ollama镜像的配�|�文�Ӟ��Modelfile

# set the base model
FROM /root/.ollama/llamafactory-export/saves/llama3-8b/lora/docker-commnad-nlp/export

# set custom parameter values
PARAMETER temperature 1
PARAMETER num_keep 24
PARAMETER stop <|start_header_id|>
PARAMETER stop <|end_header_id|>
PARAMETER stop <|eot_id|>
PARAMETER stop <|reserved_special_token

# set the model template
TEMPLATE """
{{ if .System }}<|start_header_id|>system<|end_header_id|>
{{ .System }}<|eot_id|>{{ end }}{{ if .Prompt }}<|start_header_id|>user<|end_header_id|>
{{ .Prompt }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|>
{{ .Response }}<|eot_id|>
"""

# set the system message
SYSTEM You are llama3 from Meta, customized and hosted @ Paul Wong (http://paulwong88.tpddns.cn).

# set Chinese lora support
#ADAPTER /root/.ollama/models/lora/ggml-adapter-model.bin

建立镜像命��o�Q�create-ollama-image-docker-command-nlp.sh

BIN_PATH=$(cd `dirname $0`; pwd)
cd $BIN_PATH/
pwd
ollama create llama3-docker-commnad-nlp:paul -f Modelfile

5.�q�行大模�?/h3>

llama3-docker-commnad-nlp:paul

paulwong 2024-07-08 19:48 发表评论

paulwong — Mon, 08 Jul 2024 10:44:00 GMT

对于象META的开源大模型�Q�如llama3�Q�由于都是用通用数据�q�行预训�l�，�Ҏ��使用其模型的公司来说�Q�可能会不适用�Q�因��大模型对公司的数据不熟悉�Q�因此引入微�?Fine-Tunning)�?/div>

通过喂给大模型大量数据，1万条��h��Q��得大模型也能对公司的数据熟悉�Q�进而用于各�U�对话场景�?/div>

1.克隆�q�安装LLAMA FACTORY库，install-llamafactory.sh

BIN_PATH=$(cd `dirname $0`; pwd)
cd $BIN_PATH/../
pwd
git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
cd LLaMA-Factory
pip install -e ".[torch,metrics,bitsandbytes,modelscope]"

2.讄��环境变量

export USE_MODELSCOPE_HUB=1 #使用modelscop模型库，非huggingface�?br />export CUDA_VISIBLE_DEVICES=0 �Q�设�|��用GPU
export HF_ENDPOINT=https://hf-mirror.com �Q�设�|�huggingface的替代地址
export MODELSCOPE_CACHE=/root/autodl-tmp/models/modelscope �Q�设�|�modelscope中的大模型保存�\�?br />export LLAMAFACTORY_HOME=/root/autodl-tmp/LLaMA-Factory

3.准备数据

#在data/dataset_info.json中加入此数据

"docker_command_NL": {
"hf_hub_url": "MattCoddity/dockerNLcommands"
},

在data目录中加入训�l�数据，MattCoddity/dockerNLcommands.json

数据格式为：

[
  {
    "input": "Give me a list of containers that have the Ubuntu image as their ancestor.",
    "instruction": "translate this sentence in docker command",
    "output": "docker ps --filter 'ancestor=ubuntu'"
  },

]

4.训练大模�?/h3>

训练的参数文�Ӟ��llama3_lora_sft_docker_command.yaml

### model
#md model id
model_name_or_path: LLM-Research/Meta-Llama-3-8B-Instruct
#huggingface model id
#model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct

### method
stage: sft
do_train: true
finetuning_type: lora
lora_target: all

### dataset
dataset: docker_command_NL
template: llama3
cutoff_len: 1024
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16

### output
output_dir: /root/autodl-tmp/my-test/saves/llama3-8b/lora/sft/docker-commnad-nlp/sft
logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true

### train
per_device_train_batch_size: 4
gradient_accumulation_steps: 8
learning_rate: 1.0e-4
num_train_epochs: 3.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000

### eval
val_size: 0.1
per_device_eval_batch_size: 1
eval_strategy: steps
eval_steps: 500

训练命��o�Q�lora-train-docker-command.sh

BIN_PATH=$(cd `dirname $0`; pwd)
cd $BIN_PATH/
pwd
cd $LLAMAFACTORY_HOME
pwd
llamafactory-cli train $BIN_PATH/conf/llama3_lora_sft_docker_command.yaml

执行此命令即可开始训�l�大模型�?/div>

5.合�ƈ大模�?/h3>

合�ƈ用的参数文�g�Q�llama3_lora_export_docker_command.yaml

### model
#md model id
model_name_or_path: LLM-Research/Meta-Llama-3-8B-Instruct
#huggingface model id
#model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct

adapter_name_or_path: /root/autodl-tmp/my-test/saves/llama3-8b/lora/docker-commnad-nlp/sft
template: llama3
export_dir: /root/autodl-tmp/my-test/saves/llama3-8b/lora/docker-commnad-nlp/export
finetuning_type: lora
export_size: 2
export_device: gpu
export_legacy_format: False

合�ƈ命��o�Q�lora-export-docker-command.sh

BIN_PATH=$(cd `dirname $0`; pwd)
cd $BIN_PATH/
pwd
llamafactory-cli export conf/llama3_lora_export_docker_command.yaml

paulwong 2024-07-08 18:44 发表评论

部��vdocker版的人工��OPEN-WEBUI+OLLAMA+NGINX

paulwong — Wed, 19 Jun 2024 14:23:00 GMT

一键部�|��h工智能中的OPEN-WEBUI,OLLAMA, NGINX�Q�也��对�c�M��OPEN-AI的对话机器�h

docker-compose.yaml

services:

  # ollama:
  #   deploy:
  #     resources:
  #       reservations:
  #         devices:
  #           - driver: nvidia
  #             count: all
  #             capabilities:
  #               - gpu  #使用GPU加�?br />   #   volumes:
  #     - ollama-volume:/root/.ollama #配置OLLAMA的配�|�数据文件在宿主�?br />   #     - /etc/localtime:/etc/localtime:ro
  #   container_name: ollama
  #   image: ollama/ollama
  #   restart: unless-stopped
  #   networks:
  #     - isolated #使用DOCKER的隔��ȝ��l?br />   #     - internet

  vllm:
    container_name: vllm
    image: vllm/vllm-openai:latest
    # ipc: host
    volumes:
      - ${HUGGINGFACE_MODELS_DIR}:/models
      - /etc/localtime:/etc/localtime:ro
    command: >
      --model /models/models--unsloth--llama-3-8b-Instruct-lawdata
      --served-model-name llama-3-8b-Instruct-lawdata
      --gpu-memory-utilization 0.90
      --max_model_len 1072
      --quantization bitsandbytes
      --load_format bitsandbytes
    ports:
      - "8000:8000"
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    networks:
      - isolated #使用DOCKER的隔��ȝ��l?br />
  # https://github.com/open-webui/open-webui
  open-webui: #全局�l�一的服务名
    volumes:
      - open-webui-volume:/app/backend/data #配置open-webui的配�|�数据文件在宿主�?br />       - /etc/localtime:/etc/localtime:ro
    container_name: open-webui
    restart: unless-stopped
    image: ghcr.io/open-webui/open-webui:main
    # network_mode: host
    ports:
      - "3000:3000"
    environment:
      # - OLLAMA_BASE_URL=http://ollama:11434 #OPEN-WEBUI讉K��OLLAMA的地址�Q�其实就是服务名代替IP
      - ENABLE_OLLAMA_API=False
      - OPENAI_API_BASE_URL=http://vllm:8000 /v1
      - /etc/localtime:/etc/localtime:ro
      - LOG_LEVEL=DEBUG
    depends_on:
      # - ollama
      - vllm
    networks:
      - isolated

  nginx-webui:
    volumes:
      - ${NGINX_DATA_DIR}/html:/usr/share/nginx/html:ro
      - ${NGINX_DATA_DIR}/conf/nginx.conf:/etc/nginx/nginx.conf:ro
      - ${NGINX_DATA_DIR}/conf/conf.d/default.conf:/etc/nginx/conf.d/default.conf:ro
      - ${NGINX_DATA_DIR}/conf/.htpasswd:/etc/nginx/.htpasswd:ro
      - /etc/localtime:/etc/localtime:ro
      - ${NGINX_DATA_DIR}/log/access.log:/var/log/nginx/access.log
      - ${NGINX_DATA_DIR}/log/error.log:/var/log/nginx/error.log
    container_name: nginx-webui
    ports:
      - "81:81"
    image: nginx:latest
    #image: quay.io/ricardbejarano/nginx
    depends_on:
      - open-webui
    restart: unless-stopped
    networks:
      - isolated
      - internet

volumes:
  ollama-volume:
    driver: local
    driver_opts:
      type: none
      o: bind
      device: ${OLLAMA_DATA_DIR}
  open-webui-volume:
    driver: local
    driver_opts:
      type: none
      o: bind
      device: ${OPEN_WEBUI_DATA_DIR}

networks:
  isolated:
    driver: bridge
    internal: true
  internet:
    driver: bridge

nginx.conf

user  nginx;
worker_processes  auto;

error_log  /var/log/nginx/error.log warn;
pid        /var/run/nginx.pid;

events {
    worker_connections  1024;
}

http {
    include       /etc/nginx/mime.types;
    default_type  application/octet-stream;

    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" '
                      '$status $body_bytes_sent "$http_referer" '
                      '"$http_user_agent" "$http_x_forwarded_for"';

    access_log  /var/log/nginx/access.log  main;

    sendfile        on;
    keepalive_timeout  65;

    include /etc/nginx/conf.d/*.conf;  # 加蝲 conf.d 目录下的配置文�g
}

docker/docker-nginx/data/conf/conf.d/default.conf

# server {
#     listen       80;
#     server_name  example.com www.example.com;

#     root   /usr/share/nginx/html;
#     index  index.html index.htm;

#     location / {
#         try_files $uri $uri/ =404;
#     }

#     error_page   500 502 503 504  /50x.html;
#     location = /50x.html {
#         root   /usr/share/nginx/html;
#     }
# }
server {
    listen 81;
    server_name localhost;

    location / {
        proxy_pass http://open-webui:8080;
        # proxy_pass http://localhost:8080;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
    }

    # 代理 WebSocket ��h��
    location /ws/ {
        proxy_pass http://open-webui:8080;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "Upgrade";
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
    }

    access_log /var/log/nginx/access.log;
    error_log /var/log/nginx/error.log;
}

00_varible.sh

#!/bin/bash

# 获取当前脚本的�\�?br /> # SCRIPT_PATH="$(realpath "$0")"
# echo "当前脚本的�\径是: $SCRIPT_PATH"

# 获取当前脚本所在的目录
# SCRIPT_DIR="$(dirname "$SCRIPT_PATH")"
# echo "当前脚本所在的目录�? $SCRIPT_DIR"
# cd $SCRIPT_DIR

# export HTTP_PROXY=http://192.168.0.102:7890
# export HTTPS_PROXY=https://192.168.0.102:7890

export DOCKER_ROOT_DIR=/home/paul/paulwong/work/workspaces/python-ai-project/docker
export NGINX_DATA_DIR=${DOCKER_ROOT_DIR}/docker-nginx/data
export OLLAMA_DATA_DIR=${DOCKER_ROOT_DIR}/docker-ollama/data
export OPEN_WEBUI_DATA_DIR=${DOCKER_ROOT_DIR}/docker-webui/data
export HUGGINGFACE_MODELS_DIR=/home/paul/.cache/huggingface/models

01_start-nginx-ollama-webui.sh

#!/bin/bash

# 获取当前脚本的�\�?br /> SCRIPT_PATH="$(realpath "$0")"
echo "当前脚本的�\径是: $SCRIPT_PATH"

# 获取当前脚本所在的目录
SCRIPT_DIR="$(dirname "$SCRIPT_PATH")"
echo "当前脚本所在的目录�? $SCRIPT_DIR"
cd $SCRIPT_DIR

source ./00_varible.sh
docker compose -f configs/docker-compose.yaml down
docker compose -f configs/docker-compose.yaml up

02_restart-nginx-ollama-webui.sh

03_login_ollama.sh

04_restart_open_webui.sh

paulwong 2024-06-19 22:23 发表评论