@@ -135,13 +172,17 @@ cd build/bin
Hi! I'm DeepSeek-R1, an artificial intelligence assistant created by DeepSeek. I'm at your service and would be delighted to assist you with any inquiries or tasks you may have.
```
-### GGUF Benchmark Test
+### Benchmark the model
+
+
```bash
./llama-bench -m DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M.gguf
```
-```bash
+
+
+```txt
radxa@orion-o6:~/llama.cpp/build/bin$ ./llama-bench -m ~/DeepSeek-R1-Distill-Qwen-1.5B-Q4_K_M.gguf -t 8
| model | size | params | backend | threads | test | t/s |
| ------------------------------ | ---------: | ---------: | ---------- | ------: | ------------: | -------------------: |
@@ -151,4 +192,4 @@ radxa@orion-o6:~/llama.cpp/build/bin$ ./llama-bench -m ~/DeepSeek-R1-Distill-Qwe
## References
-For more details on llama.cpp, please refer to the [official documentation](https://github.com/ggml-org/llama.cpp).
+For more details about llama.cpp, refer to the [official documentation](https://github.com/ggml-org/llama.cpp).
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_ollama.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_ollama.mdx
index 7ccac7592..d43bc9a34 100644
--- a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_ollama.mdx
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_ollama.mdx
@@ -1,27 +1,27 @@
-Ollama is a tool for running and managing large language models (LLMs) locally.
-It allows you to easily pull, run, and manage various AI models such as LLaMA, Mistral, and Gemma on your local device without complex environment configurations.
+Ollama is an efficient tool for managing and running local large language models (LLMs).
+It greatly simplifies AI model deployment. With minimal environment setup, you can pull, run, and manage models on your local device.
-## Ollama Installation
+## Install Ollama
```bash
curl -fsSL https://ollama.com/install.sh | sh
```
-For local build methods, please refer to the [official documentation](https://github.com/ollama/ollama/blob/main/docs/development.md).
+For building from source locally, refer to the [official documentation](https://github.com/ollama/ollama/blob/main/docs/development.md).
## Usage
-### Pull a Model
+### Pull a model
-This command downloads the model files from the internet.
+This command downloads the model files from the Internet.
```bash
ollama pull deepseek-r1:1.5b
```
-### Run a Model
+### Run a model
-This command runs the model directly. If the model is not cached locally, it will be downloaded automatically before running.
+This command starts the model. If it is not cached locally, Ollama will download it automatically and then run it.
```bash
ollama run deepseek-r1:1.5b
@@ -33,19 +33,19 @@ ollama run deepseek-r1:1.5b
ollama show deepseek-r1:1.5b
```
-### List models on your computer
+### List downloaded models
```bash
ollama list
```
-### List which models are currently loaded
+### List loaded models
```bash
ollama ps
```
-### Stop a model which is currently running
+### Stop a running model
```bash
ollama stop deepseek-r1:1.5b
@@ -57,6 +57,6 @@ ollama stop deepseek-r1:1.5b
ollama rm deepseek-r1:1.5b
```
-## Reference Information
+## References
-For more detailed information about Ollama, please refer to the [official documentation](https://github.com/ollama/ollama).
+For more details about Ollama, refer to the [official documentation](https://github.com/ollama/ollama).
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/accessory-use/_heatsink-8240b.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/accessory-use/_heatsink-8240b.mdx
index 7a643b400..93debc922 100644
--- a/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/accessory-use/_heatsink-8240b.mdx
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/accessory-use/_heatsink-8240b.mdx
@@ -2,9 +2,9 @@ This guide explains how to install and remove the Radxa 8420B active cooling sys
:::tip
-This tutorial is applicable to Radxa O6 / O6N. The installation and removal procedures are essentially the same for both models, with the exception that the Radxa O6 requires the additional step of removing an acrylic case.
+This tutorial is applicable to Radxa Orion O6 / O6N. The installation and removal procedures are essentially the same for both models, with the exception that the Radxa Orion O6 requires the additional step of removing an acrylic case.
-For Radxa O6N, you can follow the installation and removal instructions after removing the acrylic case.
+For Radxa Orion O6N, you can follow the installation and removal instructions after removing the acrylic case.
:::
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_API-manual.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_API-manual.mdx
new file mode 100644
index 000000000..b0897eab7
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_API-manual.mdx
@@ -0,0 +1,5 @@
+The NPU driver exposes APIs to upper-layer applications in two parts: C++ and Python.
+
+The detailed API manual can be downloaded from the [CIX Developer Center](https://developer.cixtech.com/).
+
+Scroll down to find the documentation resources, then click Download under the AI section.
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_BEV_RoadSeg.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_BEV_RoadSeg.mdx
new file mode 100644
index 000000000..fd07f4675
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_BEV_RoadSeg.mdx
@@ -0,0 +1,4 @@
+BEV_RoadSeg is a specialized system focused on drivable-area perception for autonomous driving. It combines bird’s-eye-view (BEV) transformation with a Transformer-based architecture, and uses the LSTR deep learning model to segment road structures accurately, enabling stable and reliable detection of lanes and drivable regions in complex, dynamic driving environments.
+
+- Core capability: Generates high-precision BEV segmentation maps of drivable areas and lane lines from multi-camera surround-view inputs, providing key perception signals for path planning.
+- Technical highlights: Uses the LSTR model as the core and leverages the Transformer’s strong ability to model long-range spatial relationships, effectively handling challenging scenarios such as curves, intersections, and partial occlusions.
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_ai-hub.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_ai-hub.mdx
index 9321198c8..38e536ae6 100644
--- a/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_ai-hub.mdx
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_ai-hub.mdx
@@ -1,16 +1,47 @@
-The CIX AI Model Hub is a collection of machine learning models optimized for deployment on CIX SOC. It includes AI model examples across various domains (such as computer vision, speech recognition, generative AI, and other open-source models) along with configuration files compiled for the CIX SOC NPU. This document primarily introduces how to download and run models from the AI Model Hub.
+The CIX AI Model Hub repository is hosted on ModelScope. Visit: [cix ai_model_hub](https://modelscope.cn/models/cix/ai_model_hub_25_Q3)
-## Download CIX AI Model Hub Repository
+## Clone the repository
-The CIX AI Model Hub repository is hosted on the ModelScope community platform. Access it at [cix ai_model_hub](https://modelscope.cn/models/cix/ai_model_hub_25_Q3).
+With the following command, you can clone only the files that are not tracked by Git LFS (make sure Git LFS is installed):
-You can download it using git (ensure git-lfs is installed):
+:::info
+After cloning the directory structure on the device, decide whether you also need to clone on the host depending on whether you plan to convert models on the host.
+:::
```bash
+GIT_LFS_SKIP_SMUDGE=1 git clone https://www.modelscope.cn/cix/ai_model_hub_25_Q3.git
+```
+
+## Set up the environment on the device
+
+:::info
+Activate the virtual environment before running.
+:::
+
+
+
+```bash
+python3 -m venv --system-site-packages .venv
+pip3 install -r requirements.txt
+```
+
+
+
+## Download the entire repository
+
+:::tip Recommendation
+Since the repository is large, it is recommended to avoid cloning everything.
+:::
+
+Download with Git (make sure Git LFS is installed):
+
+```bash
+mkdir ai-model-hub && cd ai-model-hub
+git lfs install
git clone https://www.modelscope.cn/cix/ai_model_hub_25_Q3.git
```
-Model example directory structure:
+Directory structure:
```bash
.
@@ -68,49 +99,3 @@ Model example directory structure:
├── text_process.py
└── tools.py
```
-
-## Running Models
-
-### Configuring the Environment
-
-Navigate to the model directory:
-
-```bash
-cd ai_model_hub_25_Q3
-```
-
-Create a Python virtual environment:
-
-```bash
-python3 -m venv venv
-```
-
-Activate the virtual environment:
-
-```bash
-source venv/bin/activate
-```
-
-Install the Python environment:
-
-```bash
-pip3 install -r requirements.txt
-```
-
-### Model Examples
-
-1. Preprocess human-readable input into model input.
-2. Run model inference.
-3. Postprocess model output into a human-readable format.
-
-All model example codes can run end-to-end on the NPU of the O6/O6N:
-
-```bash
-python3 inference_npu.py
-```
-
-Additionally, you can run the end-to-end examples locally on an X86 host or on the O6/O6N using CPU with OnnxRuntime:
-
-```bash
-python3 inference_onnx.py
-```
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_clip.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_clip.mdx
new file mode 100644
index 000000000..4bbd51480
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/orion-common/app-dev/artificial-intelligence/_clip.mdx
@@ -0,0 +1,228 @@
+**CLIP** is a general-purpose multimodal pre-trained model developed by OpenAI. By performing contrastive learning on hundreds of millions of image-text pairs collected from the Internet, it breaks away from the limitations of traditional vision models that rely on manually labeled categories, enabling AI to “understand” the visual world directly through natural language.
+
+- Key features: Strong cross-modal alignment and zero-shot transfer capability. It can recognize object categories it has never seen without task-specific fine-tuning. It is widely used for semantic image-text retrieval, automatic prompt generation, and as the core text encoder for generative AI such as Stable Diffusion.
+- Version notes: This example uses the CLIP-ViT-B/32 model. As a baseline that balances performance and deployment efficiency, it uses a Vision Transformer (ViT) as the visual backbone and processes image features with 32x32 patches. While maintaining strong semantic alignment accuracy, it has a smaller parameter size and faster inference, making it a common balanced choice for real-world multimodal applications.
+
+:::info[Environment setup]
+You need to set up the environment in advance.
+
+- [Environment setup](../../../../orion/o6/app-development/artificial-intelligence/env-setup.md)
+- [AI Model Hub](../../../../orion/o6/app-development/artificial-intelligence/ai-hub.md)
+ :::
+
+## Quick start
+
+### Download model files
+
+
+
+```bash
+cd ai_model_hub_25_Q3/models/Generative_AI/Image_to_Text/onnx_clip
+wget https://www.modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/Generative_AI/Image_to_Text/onnx_clip/clip_txt.cix
+wget https://www.modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/Generative_AI/Image_to_Text/onnx_clip/clip_visual.cix
+```
+
+
+
+### Test the model
+
+:::info
+Activate the virtual environment before running.
+:::
+
+
+
+```bash
+python3 inference_npu.py
+```
+
+
+
+## Full conversion workflow
+
+### Download model files
+
+
+
+```bash
+cd ai_model_hub_25_Q3/models/Generative_AI/Image_to_Text/onnx_clip/model
+wget https://www.modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/Generative_AI/Image_to_Text/onnx_clip/model/clip_text_model_vitb32.onnx
+wget https://www.modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/Generative_AI/Image_to_Text/onnx_clip/model/clip_visual.onnx
+```
+
+
+
+### Project structure
+
+```txt
+├── cfg
+├── clip_visual.cix
+├── clip_txt.cix
+├── datasets
+├── inference_npu.py
+├── inference_onnx.py
+├── model
+├── ReadMe.md
+└── test_data
+```
+
+### Quantize and convert the model
+
+#### Convert the image module
+
+
+
+```bash
+cd ..
+cixbuild cfg/clip_visualbuild.cfg
+```
+
+
+
+#### Convert the text module
+
+
+
+```bash
+cixbuild cfg/clip_text_model_vitb32build.cfg
+```
+
+
+
+:::info[Copy to device]
+After conversion, copy the `.cix` model files to the device.
+:::
+
+### Test inference on the host
+
+#### Run the inference script
+
+
+
+```bash
+python3 inference_onnx.py
+```
+
+
+
+#### Inference output
+
+
+
+```bash
+$ python3 inference_onnx.py
+[[0.03632354 0.96057177 0.00310465]]
+test_data/000000464522.jpg, max similarity: a dog
+[[0.03074941 0.00429748 0.9649532 ]]
+test_data/000000032811.jpg, max similarity: a bird
+[[0.8280978 0.08798673 0.08391542]]
+test_data/000000010698.jpg, max similarity: a person
+```
+
+
+
+#### Test images
+
+
+
+{" "}
+
+
+
+
+
+
+
+
+
+
+{" "}
+
+
+
+
+
+
+
+{" "}
+
+
+
+
+
+{" "}
+
+
+
+
+
+{" "}
+
+
+
+
+
+{" "}
+
+
+
+
+
+{/* Left container: portrait image */}
+
+{" "}
+
+
+
+
+
+{/* Right container: two landscape images */}
+
+{" "}
+
+
+
+
+
+
+
+
+
+
+
+
+{/* Left container: portrait image */}
+
+{" "}
+
+
+
+
+
+{/* Right container: two landscape images */}
+
+{" "}
+
+
+
+
+
+
+
+
+
+
+
+
+
+ Model test input
+
+
+
+{" "}
-Results are saved in the `output` folder
+
+
-
+
-The inference results are consistent between NPU and CPU, but the running speed is significantly faster on NPU
+{" "}
-## Reference Documents
+
+
-Paper: [Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose](https://arxiv.org/abs/1811.12004)
+
+
+
+ Model test input
+
+
+
+
+ Model test input
+
+
+
+{" "}
+
+
+
+
+
+{" "}
+
+
+
+
+
+{" "}
+
+
+
+
+
+
+{" "}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+{" "}
-Results are saved in the `output_npu` folder
+
-
+
-The inference results are consistent between NPU and CPU, but the running speed is significantly faster on NPU
+{" "}
-## Reference Documents
+
-Paper: [Accurate Image Super-Resolution Using Very Deep Convolutional Networks](https://arxiv.org/abs/1511.04587)
+
+
+
+
+
+
+
+
+{" "}
+
+
+
+
+
+
+{" "}
+
+
+
+
+
+ 
+ 
+
+ 
+ 
+
+
+
+
+
+
+
+
+