diff --git a/src/.vuepress/public/img/LargeModel01.png b/src/.vuepress/public/img/LargeModel01.png new file mode 100644 index 000000000..d5f6287c9 Binary files /dev/null and b/src/.vuepress/public/img/LargeModel01.png differ diff --git a/src/.vuepress/public/img/LargeModel02.png b/src/.vuepress/public/img/LargeModel02.png new file mode 100644 index 000000000..51c79cd29 Binary files /dev/null and b/src/.vuepress/public/img/LargeModel02.png differ diff --git a/src/.vuepress/public/img/LargeModel03.png b/src/.vuepress/public/img/LargeModel03.png new file mode 100644 index 000000000..880048bf3 Binary files /dev/null and b/src/.vuepress/public/img/LargeModel03.png differ diff --git a/src/.vuepress/public/img/LargeModel04.png b/src/.vuepress/public/img/LargeModel04.png new file mode 100644 index 000000000..c54c56a0c Binary files /dev/null and b/src/.vuepress/public/img/LargeModel04.png differ diff --git a/src/.vuepress/public/img/LargeModel05.png b/src/.vuepress/public/img/LargeModel05.png new file mode 100644 index 000000000..9ad51570c Binary files /dev/null and b/src/.vuepress/public/img/LargeModel05.png differ diff --git a/src/.vuepress/public/img/LargeModel06.png b/src/.vuepress/public/img/LargeModel06.png new file mode 100644 index 000000000..703d6b5ff Binary files /dev/null and b/src/.vuepress/public/img/LargeModel06.png differ diff --git a/src/.vuepress/public/img/LargeModel07.png b/src/.vuepress/public/img/LargeModel07.png new file mode 100644 index 000000000..c2f4e620d Binary files /dev/null and b/src/.vuepress/public/img/LargeModel07.png differ diff --git a/src/.vuepress/public/img/LargeModel08.png b/src/.vuepress/public/img/LargeModel08.png new file mode 100644 index 000000000..9d2bfab04 Binary files /dev/null and b/src/.vuepress/public/img/LargeModel08.png differ diff --git a/src/.vuepress/sidebar/V2.0.x/en-Tree.ts b/src/.vuepress/sidebar/V2.0.x/en-Tree.ts index 9802803f3..6204a9814 100644 --- a/src/.vuepress/sidebar/V2.0.x/en-Tree.ts +++ b/src/.vuepress/sidebar/V2.0.x/en-Tree.ts @@ -99,7 +99,6 @@ export const enSidebar = { children: [ { text: 'Data Sync', link: 'Data-Sync_apache' }, { text: 'Data Subscription', link: 'Data-subscription' }, - { text: 'AI Capability', link: 'AINode_apache' }, { text: 'Security Management', collapsible: true, @@ -129,6 +128,15 @@ export const enSidebar = { }, ], }, + { + text: 'AI capability', + collapsible: true, + prefix: 'AI-capability/', + children: [ + { text: 'AINode', link: 'AINode_apache' }, + { text: 'TimeSeries Large Model', link: 'TimeSeries-Large-Model' }, + ], + }, { text: 'Tools System', collapsible: true, diff --git a/src/.vuepress/sidebar/V2.0.x/zh-Tree.ts b/src/.vuepress/sidebar/V2.0.x/zh-Tree.ts index 84b92e777..c60927e5e 100644 --- a/src/.vuepress/sidebar/V2.0.x/zh-Tree.ts +++ b/src/.vuepress/sidebar/V2.0.x/zh-Tree.ts @@ -88,7 +88,6 @@ export const zhSidebar = { children: [ { text: '数据同步', link: 'Data-Sync_apache' }, { text: '数据订阅', link: 'Data-subscription' }, - { text: 'AI能力', link: 'AINode_apache' }, { text: '安全管理', collapsible: true, @@ -116,6 +115,15 @@ export const zhSidebar = { }, ], }, + { + text: 'AI 能力', + collapsible: true, + prefix: 'AI-capability/', + children: [ + { text: 'AINode', link: 'AINode_apache' }, + { text: '时序大模型', link: 'TimeSeries-Large-Model' }, + ], + }, { text: '工具体系', collapsible: true, diff --git a/src/.vuepress/sidebar_timecho/V2.0.x/en-Tree.ts b/src/.vuepress/sidebar_timecho/V2.0.x/en-Tree.ts index 318e22315..a82c0bf09 100644 --- a/src/.vuepress/sidebar_timecho/V2.0.x/en-Tree.ts +++ b/src/.vuepress/sidebar_timecho/V2.0.x/en-Tree.ts @@ -108,7 +108,6 @@ export const enSidebar = { children: [ { text: 'Data Sync', link: 'Data-Sync_timecho' }, { text: 'Data Subscription', link: 'Data-subscription' }, - { text: 'AI Capability', link: 'AINode_timecho' }, { text: 'Security Management', collapsible: true, @@ -142,6 +141,15 @@ export const enSidebar = { }, ], }, + { + text: 'AI capability', + collapsible: true, + prefix: 'AI-capability/', + children: [ + { text: 'AINode', link: 'AINode_timecho' }, + { text: 'TimeSeries Large Model', link: 'TimeSeries-Large-Model' }, + ], + }, { text: 'Tools System', collapsible: true, diff --git a/src/.vuepress/sidebar_timecho/V2.0.x/zh-Tree.ts b/src/.vuepress/sidebar_timecho/V2.0.x/zh-Tree.ts index 65a9dc222..7ea80be20 100644 --- a/src/.vuepress/sidebar_timecho/V2.0.x/zh-Tree.ts +++ b/src/.vuepress/sidebar_timecho/V2.0.x/zh-Tree.ts @@ -91,7 +91,6 @@ export const zhSidebar = { children: [ { text: '数据同步', link: 'Data-Sync_timecho' }, { text: '数据订阅', link: 'Data-subscription' }, - { text: 'AI能力', link: 'AINode_timecho' }, { text: '安全管理', collapsible: true, @@ -125,6 +124,15 @@ export const zhSidebar = { }, ], }, + { + text: 'AI 能力', + collapsible: true, + prefix: 'AI-capability/', + children: [ + { text: 'AINode', link: 'AINode_timecho' }, + { text: '时序大模型', link: 'TimeSeries-Large-Model' }, + ], + }, { text: '工具体系', collapsible: true, diff --git a/src/UserGuide/Master/Tree/User-Manual/AINode_apache.md b/src/UserGuide/Master/Tree/AI-capability/AINode_apache.md similarity index 99% rename from src/UserGuide/Master/Tree/User-Manual/AINode_apache.md rename to src/UserGuide/Master/Tree/AI-capability/AINode_apache.md index 2abf23e73..95fcccd35 100644 --- a/src/UserGuide/Master/Tree/User-Manual/AINode_apache.md +++ b/src/UserGuide/Master/Tree/AI-capability/AINode_apache.md @@ -19,7 +19,7 @@ --> -# AI Capability +# AINode AINode is the third internal node after ConfigNode and DataNode in Apache IoTDB, which extends the capability of machine learning analysis of time series by interacting with DataNode and ConfigNode of IoTDB cluster, supports the introduction of pre-existing machine learning models from the outside to be registered, and uses the registered models in the It supports the process of introducing existing machine learning models from outside for registration, and using the registered models to complete the time series analysis tasks on the specified time series data through simple SQL statements, which integrates the model creation, management and inference in the database engine. At present, we have provided machine learning algorithms or self-developed models for common timing analysis scenarios (e.g. prediction and anomaly detection). diff --git a/src/UserGuide/Master/Tree/User-Manual/AINode_timecho.md b/src/UserGuide/Master/Tree/AI-capability/AINode_timecho.md similarity index 97% rename from src/UserGuide/Master/Tree/User-Manual/AINode_timecho.md rename to src/UserGuide/Master/Tree/AI-capability/AINode_timecho.md index 793136b1e..204f91cf9 100644 --- a/src/UserGuide/Master/Tree/User-Manual/AINode_timecho.md +++ b/src/UserGuide/Master/Tree/AI-capability/AINode_timecho.md @@ -19,7 +19,7 @@ --> -# AI Capability +# AINode AINode is the third internal node after ConfigNode and DataNode in Apache IoTDB, which extends the capability of machine learning analysis of time series by interacting with DataNode and ConfigNode of IoTDB cluster, supports the introduction of pre-existing machine learning models from the outside to be registered, and uses the registered models in the It supports the process of introducing existing machine learning models from outside for registration, and using the registered models to complete the time series analysis tasks on the specified time series data through simple SQL statements, which integrates the model creation, management and inference in the database engine. At present, we have provided machine learning algorithms or self-developed models for common timing analysis scenarios (e.g. prediction and anomaly detection). @@ -444,45 +444,11 @@ Total line number = 4 In the result set, each row's label corresponds to the output of the anomaly detection model after inputting each group of 24 rows of data. -### 4.6 AINode-Timer Model Import Steps -1. Open the IoTDB CLI console and verify that the ConfigNode, DataNode, and AINode statuses are all ​Running. -Check command: -```sql -show cluster -``` - -![](/img/ainode-timer-1.png) - -2. Model file storage path: It is recommended to place the model files in the same directory as the AINode installation package. - You may create a new folder to store model files. -3. Register the model - -Use the following SQL statement: - -```sql -create model using uri -``` - -Example (for the Timer model): +### 4.6 TimeSeries Large Models Import Steps -```sql -create model Timer using uri -``` - -Note: When importing the Timer model, the name must be "Timer" (case-sensitive), otherwise it will not be recognized by the IoTDB visualization console. - -![](/img/ainode-timer-2.png) - -4. Verify model registration success - -Check command: - -```sql -show models -``` +AINode currently supports a variety of time series large models. For deployment and usage, please refer to [TimeSeries Large Models](../AI-capability/TimeSeries-Large-Model) -![](/img/ainode-timer-3.png) ## 5. Privilege Management diff --git a/src/UserGuide/Master/Tree/AI-capability/TimeSeries-Large-Model b/src/UserGuide/Master/Tree/AI-capability/TimeSeries-Large-Model new file mode 100644 index 000000000..071414d25 --- /dev/null +++ b/src/UserGuide/Master/Tree/AI-capability/TimeSeries-Large-Model @@ -0,0 +1,117 @@ + + +# TimeSeries Large Model + +## Introduction + +A time series large model is a foundational model specifically designed for time series analysis. The IoTDB team has independently developed time series large models, which are pre-trained on massive time series data using technologies such as transformer structures. These models can understand and generate time series data across various domains and are applicable to applications like time series forecasting, anomaly detection, and time series imputation. Unlike traditional time series analysis techniques, time series large models possess the capability to extract universal features and provide technical services based on zero-shot analysis and fine-tuning for a wide range of analytical tasks. + +The team's related technologies of time series large models have been published in top international machine learning conferences. + +## Application Scenarios + +- **Time Series Forecasting**: Provides forecasting services for time series data in industrial production, natural environments, and other areas, helping users to understand future trends in advance. +- **Data Imputation**: For missing segments in time series, perform context imputation to enhance the continuity and completeness of the dataset. +- **Anomaly Detection**: Utilizing regression analysis technology, monitor time series data in real-time and provide timely warnings for potential anomalies. + +![](/img/LargeModel08.png) + +## Timer Model + +The Timer model not only demonstrates excellent few-shot generalization and multi-task adaptation capabilities but also gains a rich knowledge base through pre-training, endowing it with the universal capability to handle a variety of downstream tasks, featuring the following: + +- **Generalization**: The model can be fine-tuned using a small number of samples to achieve leading predictive performance in the industry. +- **Versatility**: The model is designed flexibly to adapt to various task requirements and supports variable input and output lengths, enabling it to play a role in various application scenarios. +- **Scalability**: As the number of model parameters increases or the scale of pre-training data expands, the model's performance continues to improve, ensuring the model can optimize its predictive effects with the growth of time and data volume. + +![](/img/LargeModel02.png) + +## Timer-XL Model + +Timer-XL is an upgraded version of Timer that further extends the network structure and achieves comprehensive breakthroughs in multiple dimensions: + +- **Long Context Support**: This model breaks through the limitations of traditional time series forecasting models, supporting the processing of thousands of tokens (equivalent to tens of thousands of time points) of input, effectively addressing the bottleneck of context length. +- **Multi-variable Forecasting Scenario Coverage**: Supports a variety of forecasting scenarios, including non-stationary time series forecasting, multi-variable prediction tasks, and predictions involving covariates, meeting diverse business needs. +- **Large-scale Industrial Time Series Dataset**: Pre-trained using a massive industrial IoT time series dataset that has a large volume, excellent quality, and rich domain characteristics, covering energy, aerospace, steel, transportation, and more. + + +## Effect Demonstration + +Time series large models can adapt to real time series data from various fields and scenarios, showing excellent processing effects in various tasks. Here are the real performances on different data: + +**Time Series Forecasting:** + +Utilizing the predictive capabilities of the time series large model, it can accurately predict the future trend of time series. As shown in the figure, the blue curve represents the predicted trend, and the red curve represents the actual trend, with the two curves highly matching. + +![](/img/LargeModel03.png) + +**Data Imputation:**： + +Using the time series large model to perform predictive imputation for missing data segments. + +![](/img/LargeModel04.png) + + +**Anomaly Detection:**： + +Utilizing the time series large model to accurately identify anomalies that deviate significantly from the normal trend. + +![](/img/LargeModel05.png) + +## Deployment Usage + +1. Open the IoTDB CLI console and verify that the ConfigNode, DataNode, and AINode statuses are all ​Running. + +Check command: + +```sql +show cluster +``` + +![](/img/ainode-timer-1.png) + +2. Model file storage path: It is recommended to place the model files in the same directory as the AINode installation package. + You may create a new folder to store model files. + +3. Register the model + +Use the following SQL statement: + +```sql +create model using uri +``` + +Example (for the Timer model): + +```sql +create model Timer using uri +``` + +4. Verify model registration success + +Check command: + +```sql +show models +``` + +![](/img/LargeModel06.png) diff --git a/src/UserGuide/latest/User-Manual/AINode_apache.md b/src/UserGuide/latest/AI-capability/AINode_apache.md similarity index 99% rename from src/UserGuide/latest/User-Manual/AINode_apache.md rename to src/UserGuide/latest/AI-capability/AINode_apache.md index 2abf23e73..95fcccd35 100644 --- a/src/UserGuide/latest/User-Manual/AINode_apache.md +++ b/src/UserGuide/latest/AI-capability/AINode_apache.md @@ -19,7 +19,7 @@ --> -# AI Capability +# AINode AINode is the third internal node after ConfigNode and DataNode in Apache IoTDB, which extends the capability of machine learning analysis of time series by interacting with DataNode and ConfigNode of IoTDB cluster, supports the introduction of pre-existing machine learning models from the outside to be registered, and uses the registered models in the It supports the process of introducing existing machine learning models from outside for registration, and using the registered models to complete the time series analysis tasks on the specified time series data through simple SQL statements, which integrates the model creation, management and inference in the database engine. At present, we have provided machine learning algorithms or self-developed models for common timing analysis scenarios (e.g. prediction and anomaly detection). diff --git a/src/UserGuide/latest/User-Manual/AINode_timecho.md b/src/UserGuide/latest/AI-capability/AINode_timecho.md similarity index 97% rename from src/UserGuide/latest/User-Manual/AINode_timecho.md rename to src/UserGuide/latest/AI-capability/AINode_timecho.md index 793136b1e..204f91cf9 100644 --- a/src/UserGuide/latest/User-Manual/AINode_timecho.md +++ b/src/UserGuide/latest/AI-capability/AINode_timecho.md @@ -19,7 +19,7 @@ --> -# AI Capability +# AINode AINode is the third internal node after ConfigNode and DataNode in Apache IoTDB, which extends the capability of machine learning analysis of time series by interacting with DataNode and ConfigNode of IoTDB cluster, supports the introduction of pre-existing machine learning models from the outside to be registered, and uses the registered models in the It supports the process of introducing existing machine learning models from outside for registration, and using the registered models to complete the time series analysis tasks on the specified time series data through simple SQL statements, which integrates the model creation, management and inference in the database engine. At present, we have provided machine learning algorithms or self-developed models for common timing analysis scenarios (e.g. prediction and anomaly detection). @@ -444,45 +444,11 @@ Total line number = 4 In the result set, each row's label corresponds to the output of the anomaly detection model after inputting each group of 24 rows of data. -### 4.6 AINode-Timer Model Import Steps -1. Open the IoTDB CLI console and verify that the ConfigNode, DataNode, and AINode statuses are all ​Running. -Check command: -```sql -show cluster -``` - -![](/img/ainode-timer-1.png) - -2. Model file storage path: It is recommended to place the model files in the same directory as the AINode installation package. - You may create a new folder to store model files. -3. Register the model - -Use the following SQL statement: - -```sql -create model using uri -``` - -Example (for the Timer model): +### 4.6 TimeSeries Large Models Import Steps -```sql -create model Timer using uri -``` - -Note: When importing the Timer model, the name must be "Timer" (case-sensitive), otherwise it will not be recognized by the IoTDB visualization console. - -![](/img/ainode-timer-2.png) - -4. Verify model registration success - -Check command: - -```sql -show models -``` +AINode currently supports a variety of time series large models. For deployment and usage, please refer to [TimeSeries Large Models](../AI-capability/TimeSeries-Large-Model) -![](/img/ainode-timer-3.png) ## 5. Privilege Management diff --git a/src/UserGuide/latest/AI-capability/TimeSeries-Large-Model b/src/UserGuide/latest/AI-capability/TimeSeries-Large-Model new file mode 100644 index 000000000..071414d25 --- /dev/null +++ b/src/UserGuide/latest/AI-capability/TimeSeries-Large-Model @@ -0,0 +1,117 @@ + + +# TimeSeries Large Model + +## Introduction + +A time series large model is a foundational model specifically designed for time series analysis. The IoTDB team has independently developed time series large models, which are pre-trained on massive time series data using technologies such as transformer structures. These models can understand and generate time series data across various domains and are applicable to applications like time series forecasting, anomaly detection, and time series imputation. Unlike traditional time series analysis techniques, time series large models possess the capability to extract universal features and provide technical services based on zero-shot analysis and fine-tuning for a wide range of analytical tasks. + +The team's related technologies of time series large models have been published in top international machine learning conferences. + +## Application Scenarios + +- **Time Series Forecasting**: Provides forecasting services for time series data in industrial production, natural environments, and other areas, helping users to understand future trends in advance. +- **Data Imputation**: For missing segments in time series, perform context imputation to enhance the continuity and completeness of the dataset. +- **Anomaly Detection**: Utilizing regression analysis technology, monitor time series data in real-time and provide timely warnings for potential anomalies. + +![](/img/LargeModel08.png) + +## Timer Model + +The Timer model not only demonstrates excellent few-shot generalization and multi-task adaptation capabilities but also gains a rich knowledge base through pre-training, endowing it with the universal capability to handle a variety of downstream tasks, featuring the following: + +- **Generalization**: The model can be fine-tuned using a small number of samples to achieve leading predictive performance in the industry. +- **Versatility**: The model is designed flexibly to adapt to various task requirements and supports variable input and output lengths, enabling it to play a role in various application scenarios. +- **Scalability**: As the number of model parameters increases or the scale of pre-training data expands, the model's performance continues to improve, ensuring the model can optimize its predictive effects with the growth of time and data volume. + +![](/img/LargeModel02.png) + +## Timer-XL Model + +Timer-XL is an upgraded version of Timer that further extends the network structure and achieves comprehensive breakthroughs in multiple dimensions: + +- **Long Context Support**: This model breaks through the limitations of traditional time series forecasting models, supporting the processing of thousands of tokens (equivalent to tens of thousands of time points) of input, effectively addressing the bottleneck of context length. +- **Multi-variable Forecasting Scenario Coverage**: Supports a variety of forecasting scenarios, including non-stationary time series forecasting, multi-variable prediction tasks, and predictions involving covariates, meeting diverse business needs. +- **Large-scale Industrial Time Series Dataset**: Pre-trained using a massive industrial IoT time series dataset that has a large volume, excellent quality, and rich domain characteristics, covering energy, aerospace, steel, transportation, and more. + + +## Effect Demonstration + +Time series large models can adapt to real time series data from various fields and scenarios, showing excellent processing effects in various tasks. Here are the real performances on different data: + +**Time Series Forecasting:** + +Utilizing the predictive capabilities of the time series large model, it can accurately predict the future trend of time series. As shown in the figure, the blue curve represents the predicted trend, and the red curve represents the actual trend, with the two curves highly matching. + +![](/img/LargeModel03.png) + +**Data Imputation:**： + +Using the time series large model to perform predictive imputation for missing data segments. + +![](/img/LargeModel04.png) + + +**Anomaly Detection:**： + +Utilizing the time series large model to accurately identify anomalies that deviate significantly from the normal trend. + +![](/img/LargeModel05.png) + +## Deployment Usage + +1. Open the IoTDB CLI console and verify that the ConfigNode, DataNode, and AINode statuses are all ​Running. + +Check command: + +```sql +show cluster +``` + +![](/img/ainode-timer-1.png) + +2. Model file storage path: It is recommended to place the model files in the same directory as the AINode installation package. + You may create a new folder to store model files. + +3. Register the model + +Use the following SQL statement: + +```sql +create model using uri +``` + +Example (for the Timer model): + +```sql +create model Timer using uri +``` + +4. Verify model registration success + +Check command: + +```sql +show models +``` + +![](/img/LargeModel06.png) diff --git a/src/zh/UserGuide/latest/User-Manual/AINode_apache.md b/src/zh/UserGuide/Master/Tree/AI-capability/AINode_apache.md similarity index 99% rename from src/zh/UserGuide/latest/User-Manual/AINode_apache.md rename to src/zh/UserGuide/Master/Tree/AI-capability/AINode_apache.md index f2244a1bc..b6620803f 100644 --- a/src/zh/UserGuide/latest/User-Manual/AINode_apache.md +++ b/src/zh/UserGuide/Master/Tree/AI-capability/AINode_apache.md @@ -19,7 +19,7 @@ --> -# AI能力 +# AINode AINode 是 IoTDB 在ConfigNode、DataNode后提供的第三种内生节点，该节点通过与 IoTDB 集群的 DataNode、ConfigNode 的交互，扩展了对时间序列进行机器学习分析的能力，支持从外部引入已有机器学习模型进行注册，并使用注册的模型在指定时序数据上通过简单 SQL 语句完成时序分析任务的过程，将模型的创建、管理及推理融合在数据库引擎中。目前已提供常见时序分析场景（例如预测与异常检测）的机器学习算法或自研模型。 diff --git a/src/zh/UserGuide/latest/User-Manual/AINode_timecho.md b/src/zh/UserGuide/Master/Tree/AI-capability/AINode_timecho.md similarity index 98% rename from src/zh/UserGuide/latest/User-Manual/AINode_timecho.md rename to src/zh/UserGuide/Master/Tree/AI-capability/AINode_timecho.md index 9bbe372cd..c67c41de5 100644 --- a/src/zh/UserGuide/latest/User-Manual/AINode_timecho.md +++ b/src/zh/UserGuide/Master/Tree/AI-capability/AINode_timecho.md @@ -19,7 +19,7 @@ --> -# AI能力 +# AINode AINode 是 IoTDB 在ConfigNode、DataNode后提供的第三种内生节点，该节点通过与 IoTDB 集群的 DataNode、ConfigNode 的交互，扩展了对时间序列进行机器学习分析的能力，支持从外部引入已有机器学习模型进行注册，并使用注册的模型在指定时序数据上通过简单 SQL 语句完成时序分析任务的过程，将模型的创建、管理及推理融合在数据库引擎中。目前已提供常见时序分析场景（例如预测与异常检测）的机器学习算法或自研模型。 @@ -441,43 +441,10 @@ Total line number = 4 其中结果集中每行的标签对应每24行数据为一组，输入该异常检测模型后的输出。 -### 4.6 AINode-Timer模型导入操作步骤 -1. 打开 IoTDB cli 控制台，检查 ConfigNode、DataNode、AINode 节点确保均为 Running。 +### 4.6 时序大模型导入步骤 -检查命令： -```sql -show cluster -``` - -![](/img/ainode-timer-1.png) - -2. 模型文件存放路径：推荐放在 AINode 安装包相同文件夹下，可新建模型文件夹存放模型文件 -3. 注册模型语句 - -```sql -create model using uri -``` - -示例： - -```sql -create model Timer using uri -``` - -注意：导入 Timer 模型时名称需命名为“Timer”，否则无法被可视化控制台识别。 - -![](/img/ainode-timer-2.png) - -4. 检查模型是否注册成功 - -检查命令： - -```sql -show models -``` - -![](/img/ainode-timer-3.png) +AINode 目前支持多种时序大模型，部署使用请参考[时序大模型](../AI-capability/TimeSeries-Large-Model) ## 5. 权限管理 diff --git a/src/zh/UserGuide/Master/Tree/AI-capability/TimeSeries-Large-Model b/src/zh/UserGuide/Master/Tree/AI-capability/TimeSeries-Large-Model new file mode 100644 index 000000000..5fe29f7c5 --- /dev/null +++ b/src/zh/UserGuide/Master/Tree/AI-capability/TimeSeries-Large-Model @@ -0,0 +1,112 @@ + + +# 时序大模型 + +## 简介 + +时序大模型是一种专为时序数据分析设计的基础模型。IoTDB 团队长期自研时序大模型，基于变换器（Transformer）结构等技术在海量时序数据上预训练，能够理解并生成多种领域的时序数据，可被应用于时序预测、异常检测、时序填补等应用场景。不同于传统时序分析技术，时序大模型具备通用特征提取能力，基于零样本分析、微调等技术服务广泛的分析任务。 + +团队所研时序大模型相关技术均发表在国际机器学习顶级会议。 + +## 应用场景 + +- **时序预测**：为工业生产、自然环境等领域提供时间序列数据的预测服务，帮助用户提前了解未来趋势。 +- **数据填补**：针对时间序列中的缺失序列段，进行上下文填补，以增强数据集的连续性和完整性。 +- **异常检测**：利用自回归分析技术，对时间序列数据进行实时监测，及时预警潜在的异常情况。 + +![](/img/LargeModel07.png) + +## Timer 模型 + +Timer模型不仅展现了出色的少样本泛化和多任务适配能力，还通过预训练获得了丰富的知识库，赋予了它处理多样化下游任务的通用能力，拥有以下特点： + +- **泛化性**：模型能够通过使用少量样本进行微调，达到行业内领先的深度模型预测效果。 +- **通用性**：模型设计灵活，能够适配多种不同的任务需求，并且支持变化的输入和输出长度，使其在各种应用场景中都能发挥作用。 +- **可扩展性**：随着模型参数数量的增加或预训练数据规模的扩大，模型的性能会持续提升，确保模型能够随着时间和数据量的增长而不断优化其预测效果。 + +![](/img/LargeModel02.png) + +## Timer-XL 模型 + +Timer-XL 基于 Timer 进一步扩展升级了网络结构，在多个维度上进行全面突破： + +- **超长上下文支持**：该模型突破了传统时序预测模型的限制，支持处理数千个Token（相当于数万个时间点）的输入，有效解决了上下文长度的瓶颈问题。 +- **多变量预测场景覆盖**：支持多种预测场景，包括非平稳时间序列的预测、涉及多个变量的预测任务以及包含协变量的预测，满足多样化的业务需求。 +- **大规模工业时序数据集：**采用万亿大规模工业物联网领域的时序数据集进行预训练，数据集兼有庞大的体量、卓越的质量和丰富的领域等重要特质，覆盖能源、航空航天、钢铁、交通等多领域。 + + +## 效果展示 + +时序大模型能够适应多种不同领域和场景的真实时序数据，在各种任务上拥有优异的处理效果，以下是在不同数据上的真实表现： + +**时序预测：** + +利用时序大模型的预测能力，能够准确预测时间序列的未来变化趋势，如下图蓝色曲线代表预测趋势，红色曲线为实际趋势，两曲线高度吻合。 + +![](/img/LargeModel03.png) + +**数据填补：**： + +利用时序大模型对缺失数据段进行预测式填补。 + +![](/img/LargeModel04.png) + + +**异常检测：**： + +利用时序大模型精准识别与正常趋势偏离过大的异常值。 + +![](/img/LargeModel05.png) + +## 部署使用 + +1. 打开 IoTDB cli 控制台，检查 ConfigNode、DataNode、AINode 节点确保均为 Running。 + +检查命令： +```sql +show cluster +``` + +![](/img/ainode-timer-1.png) + +2. 模型文件存放路径：推荐放在 AINode 安装包相同文件夹下，可新建模型文件夹存放模型文件 +3. 注册模型语句 + +```sql +create model using uri +``` + +示例： + +```sql +create model Timer-xl using uri +``` + +4. 检查模型是否注册成功 + +检查命令： + +```sql +show models +``` + +![](/img/LargeModel06.png) diff --git a/src/zh/UserGuide/Master/Tree/User-Manual/AINode_apache.md b/src/zh/UserGuide/latest/AI-capability/AINode_apache.md similarity index 99% rename from src/zh/UserGuide/Master/Tree/User-Manual/AINode_apache.md rename to src/zh/UserGuide/latest/AI-capability/AINode_apache.md index f2244a1bc..b6620803f 100644 --- a/src/zh/UserGuide/Master/Tree/User-Manual/AINode_apache.md +++ b/src/zh/UserGuide/latest/AI-capability/AINode_apache.md @@ -19,7 +19,7 @@ --> -# AI能力 +# AINode AINode 是 IoTDB 在ConfigNode、DataNode后提供的第三种内生节点，该节点通过与 IoTDB 集群的 DataNode、ConfigNode 的交互，扩展了对时间序列进行机器学习分析的能力，支持从外部引入已有机器学习模型进行注册，并使用注册的模型在指定时序数据上通过简单 SQL 语句完成时序分析任务的过程，将模型的创建、管理及推理融合在数据库引擎中。目前已提供常见时序分析场景（例如预测与异常检测）的机器学习算法或自研模型。 diff --git a/src/zh/UserGuide/Master/Tree/User-Manual/AINode_timecho.md b/src/zh/UserGuide/latest/AI-capability/AINode_timecho.md similarity index 98% rename from src/zh/UserGuide/Master/Tree/User-Manual/AINode_timecho.md rename to src/zh/UserGuide/latest/AI-capability/AINode_timecho.md index 9bbe372cd..c67c41de5 100644 --- a/src/zh/UserGuide/Master/Tree/User-Manual/AINode_timecho.md +++ b/src/zh/UserGuide/latest/AI-capability/AINode_timecho.md @@ -19,7 +19,7 @@ --> -# AI能力 +# AINode AINode 是 IoTDB 在ConfigNode、DataNode后提供的第三种内生节点，该节点通过与 IoTDB 集群的 DataNode、ConfigNode 的交互，扩展了对时间序列进行机器学习分析的能力，支持从外部引入已有机器学习模型进行注册，并使用注册的模型在指定时序数据上通过简单 SQL 语句完成时序分析任务的过程，将模型的创建、管理及推理融合在数据库引擎中。目前已提供常见时序分析场景（例如预测与异常检测）的机器学习算法或自研模型。 @@ -441,43 +441,10 @@ Total line number = 4 其中结果集中每行的标签对应每24行数据为一组，输入该异常检测模型后的输出。 -### 4.6 AINode-Timer模型导入操作步骤 -1. 打开 IoTDB cli 控制台，检查 ConfigNode、DataNode、AINode 节点确保均为 Running。 +### 4.6 时序大模型导入步骤 -检查命令： -```sql -show cluster -``` - -![](/img/ainode-timer-1.png) - -2. 模型文件存放路径：推荐放在 AINode 安装包相同文件夹下，可新建模型文件夹存放模型文件 -3. 注册模型语句 - -```sql -create model using uri -``` - -示例： - -```sql -create model Timer using uri -``` - -注意：导入 Timer 模型时名称需命名为“Timer”，否则无法被可视化控制台识别。 - -![](/img/ainode-timer-2.png) - -4. 检查模型是否注册成功 - -检查命令： - -```sql -show models -``` - -![](/img/ainode-timer-3.png) +AINode 目前支持多种时序大模型，部署使用请参考[时序大模型](../AI-capability/TimeSeries-Large-Model) ## 5. 权限管理 diff --git a/src/zh/UserGuide/latest/AI-capability/TimeSeries-Large-Model b/src/zh/UserGuide/latest/AI-capability/TimeSeries-Large-Model new file mode 100644 index 000000000..06867516c --- /dev/null +++ b/src/zh/UserGuide/latest/AI-capability/TimeSeries-Large-Model @@ -0,0 +1,111 @@ + + +# 时序大模型 + +## 简介 + +时序大模型是一种专为时序数据分析设计的基础模型。IoTDB 团队长期自研时序大模型，基于变换器（Transformer）结构等技术在海量时序数据上预训练，能够理解并生成多种领域的时序数据，可被应用于时序预测、异常检测、时序填补等应用场景。不同于传统时序分析技术，时序大模型具备通用特征提取能力，基于零样本分析、微调等技术服务广泛的分析任务。 + +团队所研时序大模型相关技术均发表在国际机器学习顶级会议。 + +## 应用场景 + +- **时序预测**：为工业生产、自然环境等领域提供时间序列数据的预测服务，帮助用户提前了解未来趋势。 +- **数据填补**：针对时间序列中的缺失序列段，进行上下文填补，以增强数据集的连续性和完整性。 +- **异常检测**：利用自回归分析技术，对时间序列数据进行实时监测，及时预警潜在的异常情况。 + +![](/img/LargeModel07.png) + +## Timer 模型 + +Timer模型不仅展现了出色的少样本泛化和多任务适配能力，还通过预训练获得了丰富的知识库，赋予了它处理多样化下游任务的通用能力，拥有以下特点： + +- **泛化性**：模型能够通过使用少量样本进行微调，达到行业内领先的深度模型预测效果。 +- **通用性**：模型设计灵活，能够适配多种不同的任务需求，并且支持变化的输入和输出长度，使其在各种应用场景中都能发挥作用。 +- **可扩展性**：随着模型参数数量的增加或预训练数据规模的扩大，模型的性能会持续提升，确保模型能够随着时间和数据量的增长而不断优化其预测效果。 + +![](/img/LargeModel02.png) + +## Timer-XL 模型 + +Timer-XL 基于 Timer 进一步扩展升级了网络结构，在多个维度上进行全面突破： + +- **超长上下文支持**：该模型突破了传统时序预测模型的限制，支持处理数千个Token（相当于数万个时间点）的输入，有效解决了上下文长度的瓶颈问题。 +- **多变量预测场景覆盖**：支持多种预测场景，包括非平稳时间序列的预测、涉及多个变量的预测任务以及包含协变量的预测，满足多样化的业务需求。 +- **大规模工业时序数据集：**采用万亿大规模工业物联网领域的时序数据集进行预训练，数据集兼有庞大的体量、卓越的质量和丰富的领域等重要特质，覆盖能源、航空航天、钢铁、交通等多领域。 + +## 效果展示 + +时序大模型能够适应多种不同领域和场景的真实时序数据，在各种任务上拥有优异的处理效果，以下是在不同数据上的真实表现： + +**时序预测：** + +利用时序大模型的预测能力，能够准确预测时间序列的未来变化趋势，如下图蓝色曲线代表预测趋势，红色曲线为实际趋势，两曲线高度吻合。 + +![](/img/LargeModel03.png) + +**数据填补：**： + +利用时序大模型对缺失数据段进行预测式填补。 + +![](/img/LargeModel04.png) + + +**异常检测：**： + +利用时序大模型精准识别与正常趋势偏离过大的异常值。 + +![](/img/LargeModel05.png) + +## 部署使用 + +1. 打开 IoTDB cli 控制台，检查 ConfigNode、DataNode、AINode 节点确保均为 Running。 + +检查命令： +```sql +show cluster +``` + +![](/img/ainode-timer-1.png) + +2. 模型文件存放路径：推荐放在 AINode 安装包相同文件夹下，可新建模型文件夹存放模型文件 +3. 注册模型语句 + +```sql +create model using uri +``` + +示例： + +```sql +create model Timer-xl using uri +``` + +4. 检查模型是否注册成功 + +检查命令： + +```sql +show models +``` + +![](/img/LargeModel06.png)