alibaba · kkkky123 · Nov 13, 2025 · Nov 13, 2025
diff --git a/...h/DesignImplementation/AgenticPipeline.md → ...s/DesignImplementation/AgenticPipeline.md b/...h/DesignImplementation/AgenticPipeline.md → ...s/DesignImplementation/AgenticPipeline.md
@@ -2,7 +2,7 @@
 
 ## Agentic Pipeline Architecture Diagram
 
-![Agentic Pipeline Architecture Diagram](../../../static/img/AgenticPipeline.jpg)
+![Agentic Pipeline Architecture Diagram](../../static/img/AgenticPipeline.jpg)
 
 ## AgenticPipeline
 

diff --git a/...lish/DesignImplementation/RLVRPipeline.md → ...docs/DesignImplementation/RLVRPipeline.md b/...lish/DesignImplementation/RLVRPipeline.md → ...docs/DesignImplementation/RLVRPipeline.md
diff --git a/...glish/DevelopmentGuide/customer_env_en.md → .../docs/DevelopmentGuide/customer_env_en.md b/...glish/DevelopmentGuide/customer_env_en.md → .../docs/DevelopmentGuide/customer_env_en.md
diff --git a/...glish/DevelopmentGuide/prompt_intro_en.md → .../docs/DevelopmentGuide/prompt_intro_en.md b/...glish/DevelopmentGuide/prompt_intro_en.md → .../docs/DevelopmentGuide/prompt_intro_en.md
diff --git a/...DevelopmentGuide/support_new_models_en.md → ...DevelopmentGuide/support_new_models_en.md b/...DevelopmentGuide/support_new_models_en.md → ...DevelopmentGuide/support_new_models_en.md
diff --git a/...l/docs/English/QuickStart/config_guide.md → docs_roll/docs/QuickStart/config_guide.md b/...l/docs/English/QuickStart/config_guide.md → docs_roll/docs/QuickStart/config_guide.md
@@ -2,7 +2,7 @@
 
 ## Pipeline Config
 
-Refer to [RLVR Pipeline Start](https://alibaba.github.io/ROLL/docs/English/StepByStep/rlvr_pipeline_start) and [Agentic Pipeline Start](https://alibaba.github.io/ROLL/docs/English/StepByStep/agent_pipeline_start) for more details about RLVR/Agentic pipeline configurations and reward settings.
+Refer to [RLVR Pipeline Start](https://alibaba.github.io/ROLL/docs/StepByStep/rlvr_pipeline_start) and [Agentic Pipeline Start](https://alibaba.github.io/ROLL/docs/StepByStep/agent_pipeline_start) for more details about RLVR/Agentic pipeline configurations and reward settings.
 
 ```yaml
 rollout_batch_size: 64

diff --git a/.../docs/English/QuickStart/config_system.md → docs_roll/docs/QuickStart/config_system.md b/.../docs/English/QuickStart/config_system.md → docs_roll/docs/QuickStart/config_system.md
@@ -58,7 +58,7 @@ To better understand ROLL's configuration system, the UML diagram of RLVRConfig
 - In WorkerConfig, ModelArguments, TrainingArguments, GeneratingArguments, DataArguments, and backend StrategyArguments held by the role are defined as needed
 - WorkerConfig can be extended as needed to construct new application configurations
 
-![RLVRConfig Diagram](../../../static/img/roll_config_diagram.png)
+![RLVRConfig Diagram](../../static/img/roll_config_diagram.png)
 
 ## Mapping between YAML Configuration and PipelineConfig
 

diff --git a/...ll/docs/English/QuickStart/debug_guide.md → docs_roll/docs/QuickStart/debug_guide.md b/...ll/docs/English/QuickStart/debug_guide.md → docs_roll/docs/QuickStart/debug_guide.md
diff --git a/.../docs/English/QuickStart/image_address.md → docs_roll/docs/QuickStart/image_address.md b/.../docs/English/QuickStart/image_address.md → docs_roll/docs/QuickStart/image_address.md
diff --git a/...l/docs/English/QuickStart/installation.md → docs_roll/docs/QuickStart/installation.md b/...l/docs/English/QuickStart/installation.md → docs_roll/docs/QuickStart/installation.md
@@ -2,7 +2,7 @@
 
 ## 🐳 Install from Docker
 
-We provide pre-built Docker images both on CUDA and ROCm for a quick start. Choose your desired image from the [Image Addresses](https://alibaba.github.io/ROLL/docs/English/QuickStart/image_address).
+We provide pre-built Docker images both on CUDA and ROCm for a quick start. Choose your desired image from the [Image Addresses](https://alibaba.github.io/ROLL/docs/QuickStart/image_address).
 
 ## 🛠️ Install from Custom Environment
 
@@ -35,7 +35,7 @@ git clone https://github.com/alibaba/ROLL.git
 cd ROLL
 pip install -r requirements.txt # Or follow your specific installation steps
 ```
-We highly suggest to use pre-built Docker images from [Image Addresses](https://alibaba.github.io/ROLL/docs/English/QuickStart/image_address) instead of installation from Custom Environment for ROCm users.
+We highly suggest to use pre-built Docker images from [Image Addresses](https://alibaba.github.io/ROLL/docs/QuickStart/image_address) instead of installation from Custom Environment for ROCm users.
 
 ## 🔄 About Model Checkpoint Format
 

diff --git a/...ish/QuickStart/multi_nodes_quick_start.md → ...ocs/QuickStart/multi_nodes_quick_start.md b/...ish/QuickStart/multi_nodes_quick_start.md → ...ocs/QuickStart/multi_nodes_quick_start.md
@@ -9,7 +9,7 @@ curl -fsSL https://github.com/alibaba/ROLL/blob/main/scripts/install_docker_nvid
 ```
 
 ## Environment Configuration
-Choose your desired Docker image from the [image addresses](https://alibaba.github.io/ROLL/docs/English/QuickStart/image_address). The following example uses *torch2.6.0 + vLLM0.8.4*
+Choose your desired Docker image from the [image addresses](https://alibaba.github.io/ROLL/docs/QuickStart/image_address). The following example uses *torch2.6.0 + vLLM0.8.4*
 ```shell
 # 1. Start a Docker container with GPU support, expose container ports, and keep the container running
 sudo docker run -dit \
@@ -56,7 +56,7 @@ Notes:
 bash examples/agentic_demo/run_agentic_pipeline_frozen_lake_multi_nodes_demo.sh
 ```
 After the Ray cluster starts, you will see log examples like the following:
-![log_ray_multi_nodes](../../../static/img/log_ray_multi_nodes.png)
+![log_ray_multi_nodes](../../static/img/log_ray_multi_nodes.png)
 
 3. Configure environment variables on the worker node
 ```shell

diff --git a/...roll/docs/English/QuickStart/qa_issues.md → docs_roll/docs/QuickStart/qa_issues.md b/...roll/docs/English/QuickStart/qa_issues.md → docs_roll/docs/QuickStart/qa_issues.md
@@ -104,7 +104,7 @@ Check the total number of GPUs requested and the `device_mapping` configuration.
 
 When ROLL distributes data, it will distribute `rollout_batch_size` samples to each `actor_train` worker according to DP size, and then calculate the samples for each gradient update according to `gradient_accumulation_steps`. The configuration results in 0 when divided.
 
-For detailed configuration logic, refer to the manual: [Training Arguments](https://alibaba.github.io/ROLL/docs/English/QuickStart/config_guide#training-arguments-training_args)
+For detailed configuration logic, refer to the manual: [Training Arguments](https://alibaba.github.io/ROLL/docs/QuickStart/config_guide#training-arguments-training_args)
 
 ### Error: `AssertionError: batch_size 32 < chunks 64`
 

diff --git a/...ish/QuickStart/single_node_quick_start.md → ...ocs/QuickStart/single_node_quick_start.md b/...ish/QuickStart/single_node_quick_start.md → ...ocs/QuickStart/single_node_quick_start.md
@@ -9,7 +9,7 @@ curl -fsSL https://github.com/alibaba/ROLL/blob/main/scripts/install_docker_nvid
 ```
 
 ## Environment Configuration
-Choose your desired Docker image from the [image addresses](https://alibaba.github.io/ROLL/docs/English/QuickStart/image_address). The following example uses *torch2.6.0 + vLLM0.8.4*
+Choose your desired Docker image from the [image addresses](https://alibaba.github.io/ROLL/docs/QuickStart/image_address). The following example uses *torch2.6.0 + vLLM0.8.4*
 ```shell
 # 1. Start a Docker container with GPU support, expose container ports, and keep the container running
 sudo docker run -dit \
@@ -41,11 +41,11 @@ bash examples/agentic_demo/run_agentic_pipeline_frozen_lake_single_node_demo.sh
 ```
 
 Example log screenshots during pipeline execution:
-![log_pipeline_start](../../../static/img/log_pipeline_start.png)
+![log_pipeline_start](../../static/img/log_pipeline_start.png)
 
-![log_pipeline_in_training](../../../static/img/log_pipeline_in_training.png)
+![log_pipeline_in_training](../../static/img/log_pipeline_in_training.png)
 
-![log_pipeline_complete](../../../static/img/log_pipeline_complete.png)
+![log_pipeline_complete](../../static/img/log_pipeline_complete.png)
 
 ## Reference: Single V100 GPU Memory Configuration Key Points
 ```yaml

diff --git a/...ocs/English/UserGuide/agentic/Tool_Use.md → docs_roll/docs/UserGuide/agentic/Tool_Use.md b/...ocs/English/UserGuide/agentic/Tool_Use.md → docs_roll/docs/UserGuide/agentic/Tool_Use.md
@@ -138,15 +138,15 @@ dapo_17k_with_python_code:
 }
 ```
 
-![math dapo + python_tool](../../../../static/img/math_python_tool.png)
+![math dapo + python_tool](../../../static/img/math_python_tool.png)
 
 ### QA + SearchTool
 
 #### Training Configuration
 - HotpotQA: `examples/qwen3_agentic_gem/gem_math_hotpotqa.yaml`
 - HotpotQA + Search: `examples/qwen3_agentic_gem/gem_math_hotpotqa_search.yaml`
 
-![qa + search](../../../../static/img/qa_search.png)
+![qa + search](../../../static/img/qa_search.png)
 
 #### Search Engine Startup Method
 

diff --git a/...nglish/UserGuide/agentic/agentic_GiGPO.md → ...l/docs/UserGuide/agentic/agentic_GiGPO.md b/...nglish/UserGuide/agentic/agentic_GiGPO.md → ...l/docs/UserGuide/agentic/agentic_GiGPO.md
diff --git a/...glish/UserGuide/agentic/agentic_StarPO.md → .../docs/UserGuide/agentic/agentic_StarPO.md b/...glish/UserGuide/agentic/agentic_StarPO.md → .../docs/UserGuide/agentic/agentic_StarPO.md
diff --git a/...uide/agentic/agentic_engineer_practice.md → ...uide/agentic/agentic_engineer_practice.md b/...uide/agentic/agentic_engineer_practice.md → ...uide/agentic/agentic_engineer_practice.md
@@ -298,4 +298,4 @@ rollout_dump_dir: odps://odps_project/tables/table_name/ds=${model_name}
 
 A: You can debug the trajectory generation process by configuring log levels and adding custom logs. Trajectory data will be completely saved in JSON format for easy analysis.
 
-For multi-round interaction local debugging, refer to the documentation: [Debug Guide](../../../简体中文/快速开始/debug_guide.md)
+For multi-round interaction local debugging, refer to the documentation: [Debug Guide](../../QuickStart/debug_guide.md)
diff --git a/...docs/English/UserGuide/algorithms/GRPO.md → docs_roll/docs/UserGuide/algorithms/GRPO.md b/...docs/English/UserGuide/algorithms/GRPO.md → docs_roll/docs/UserGuide/algorithms/GRPO.md
diff --git a/...docs/English/UserGuide/algorithms/GSPO.md → docs_roll/docs/UserGuide/algorithms/GSPO.md b/...docs/English/UserGuide/algorithms/GSPO.md → docs_roll/docs/UserGuide/algorithms/GSPO.md
diff --git a/...s/English/UserGuide/algorithms/LitePPO.md → ...roll/docs/UserGuide/algorithms/LitePPO.md b/...s/English/UserGuide/algorithms/LitePPO.md → ...roll/docs/UserGuide/algorithms/LitePPO.md
diff --git a/.../docs/English/UserGuide/algorithms/PPO.md → docs_roll/docs/UserGuide/algorithms/PPO.md b/.../docs/English/UserGuide/algorithms/PPO.md → docs_roll/docs/UserGuide/algorithms/PPO.md
diff --git a/...sh/UserGuide/algorithms/RAFT_Plus_Plus.md → ...cs/UserGuide/algorithms/RAFT_Plus_Plus.md b/...sh/UserGuide/algorithms/RAFT_Plus_Plus.md → ...cs/UserGuide/algorithms/RAFT_Plus_Plus.md
diff --git a/...erGuide/algorithms/Reinforce_Plus_Plus.md → ...erGuide/algorithms/Reinforce_Plus_Plus.md b/...erGuide/algorithms/Reinforce_Plus_Plus.md → ...erGuide/algorithms/Reinforce_Plus_Plus.md
diff --git a/...English/UserGuide/algorithms/Reward_FL.md → ...ll/docs/UserGuide/algorithms/Reward_FL.md b/...English/UserGuide/algorithms/Reward_FL.md → ...ll/docs/UserGuide/algorithms/Reward_FL.md
diff --git a/...docs/English/UserGuide/algorithms/TOPR.md → docs_roll/docs/UserGuide/algorithms/TOPR.md b/...docs/English/UserGuide/algorithms/TOPR.md → docs_roll/docs/UserGuide/algorithms/TOPR.md
diff --git a/...UserGuide/algorithms/offpolicy_setting.md → ...UserGuide/algorithms/offpolicy_setting.md b/...UserGuide/algorithms/offpolicy_setting.md → ...UserGuide/algorithms/offpolicy_setting.md
diff --git a/.../English/UserGuide/ascend/ascend_usage.md → ...oll/docs/UserGuide/ascend/ascend_usage.md b/.../English/UserGuide/ascend/ascend_usage.md → ...oll/docs/UserGuide/ascend/ascend_usage.md
diff --git a/...glish/UserGuide/async_parallel_rollout.md → .../docs/UserGuide/async_parallel_rollout.md b/...glish/UserGuide/async_parallel_rollout.md → .../docs/UserGuide/async_parallel_rollout.md
diff --git a/.../docs/English/UserGuide/async_training.md → docs_roll/docs/UserGuide/async_training.md b/.../docs/English/UserGuide/async_training.md → docs_roll/docs/UserGuide/async_training.md
@@ -188,7 +188,7 @@ actor_infer:
 - Configuration example: `examples/qwen2.5-7B-rlvr_megatron/rlvr_config_async.yaml`
 
 ### Off-Policy Algorithms
-ROLL also supports various Off-Policy algorithms. For detailed information, please refer to: `docs_roll/docs/English/UserGuide/algorithms/offpolicy_setting.md`
+ROLL also supports various Off-Policy algorithms. For detailed information, please refer to: `docs_roll/docs/UserGuide/algorithms/offpolicy_setting.md`
 
 Configuration example: `examples/qwen2.5-7B-rlvr-offpolicy/rlvr_config.yaml`
 

diff --git a/...cs/English/UserGuide/backend/deepspeed.md → ..._roll/docs/UserGuide/backend/deepspeed.md b/...cs/English/UserGuide/backend/deepspeed.md → ..._roll/docs/UserGuide/backend/deepspeed.md
diff --git a/.../English/UserGuide/backend/fp8_rollout.md → ...oll/docs/UserGuide/backend/fp8_rollout.md b/.../English/UserGuide/backend/fp8_rollout.md → ...oll/docs/UserGuide/backend/fp8_rollout.md
diff --git a/...ll/docs/English/UserGuide/backend/lora.md → docs_roll/docs/UserGuide/backend/lora.md b/...ll/docs/English/UserGuide/backend/lora.md → docs_roll/docs/UserGuide/backend/lora.md
diff --git a/...ocs/English/UserGuide/backend/megatron.md → docs_roll/docs/UserGuide/backend/megatron.md b/...ocs/English/UserGuide/backend/megatron.md → docs_roll/docs/UserGuide/backend/megatron.md
diff --git a/.../docs/English/UserGuide/backend/sglang.md → docs_roll/docs/UserGuide/backend/sglang.md b/.../docs/English/UserGuide/backend/sglang.md → docs_roll/docs/UserGuide/backend/sglang.md
diff --git a/...ll/docs/English/UserGuide/backend/vllm.md → docs_roll/docs/UserGuide/backend/vllm.md b/...ll/docs/English/UserGuide/backend/vllm.md → docs_roll/docs/UserGuide/backend/vllm.md
diff --git a/...nglish/UserGuide/checkpoint_and_resume.md → ...l/docs/UserGuide/checkpoint_and_resume.md b/...nglish/UserGuide/checkpoint_and_resume.md → ...l/docs/UserGuide/checkpoint_and_resume.md
diff --git a/.../docs/English/UserGuide/device_mapping.md → docs_roll/docs/UserGuide/device_mapping.md b/.../docs/English/UserGuide/device_mapping.md → docs_roll/docs/UserGuide/device_mapping.md
diff --git a/...nglish/UserGuide/megatron_convert_2_hf.md → ...l/docs/UserGuide/megatron_convert_2_hf.md b/...nglish/UserGuide/megatron_convert_2_hf.md → ...l/docs/UserGuide/megatron_convert_2_hf.md
diff --git a/...glish/UserGuide/offload_reload_control.md → .../docs/UserGuide/offload_reload_control.md b/...glish/UserGuide/offload_reload_control.md → .../docs/UserGuide/offload_reload_control.md
diff --git a/...serGuide/pipeline/agent_pipeline_start.md → ...serGuide/pipeline/agent_pipeline_start.md b/...serGuide/pipeline/agent_pipeline_start.md → ...serGuide/pipeline/agent_pipeline_start.md
diff --git a/...rGuide/pipeline/agentic_pipeline_start.md → ...rGuide/pipeline/agentic_pipeline_start.md b/...rGuide/pipeline/agentic_pipeline_start.md → ...rGuide/pipeline/agentic_pipeline_start.md
@@ -243,7 +243,7 @@ python examples/start_agentic_pipeline.py \
 
 ### Step 5: Output and Results
 
-* **Trained Model** – Checkpoints are saved in `checkpoint_config`, refer to documentation [checkpoint_and_resume](././checkpoint_and_resume.md) for details.
+* **Trained Model** – Checkpoints are saved in `checkpoint_config`, refer to documentation [checkpoint_and_resume](../../UserGuide/checkpoint_and_resume) for details.
 * **Evaluation Metrics** – Recorded in TensorBoard and terminal.
 * **Rendered Frames** – If `render_save_dir` is configured, environment rendered frames will be saved in that directory, facilitating visualization of the interaction process.
 

diff --git a/...rGuide/pipeline/distill_pipeline_start.md → ...rGuide/pipeline/distill_pipeline_start.md b/...rGuide/pipeline/distill_pipeline_start.md → ...rGuide/pipeline/distill_pipeline_start.md
diff --git a/.../UserGuide/pipeline/dpo_pipeline_start.md → .../UserGuide/pipeline/dpo_pipeline_start.md b/.../UserGuide/pipeline/dpo_pipeline_start.md → .../UserGuide/pipeline/dpo_pipeline_start.md
diff --git a/...UserGuide/pipeline/rlvr_pipeline_start.md → ...UserGuide/pipeline/rlvr_pipeline_start.md b/...UserGuide/pipeline/rlvr_pipeline_start.md → ...UserGuide/pipeline/rlvr_pipeline_start.md
diff --git a/...rGuide/pipeline/vl_rlvr_pipeline_start.md → ...rGuide/pipeline/vl_rlvr_pipeline_start.md b/...rGuide/pipeline/vl_rlvr_pipeline_start.md → ...rGuide/pipeline/vl_rlvr_pipeline_start.md
diff --git a/docs_roll/docs/English/start.mdx → docs_roll/docs/UserGuide/start.mdx b/docs_roll/docs/English/start.mdx → docs_roll/docs/UserGuide/start.mdx
@@ -42,15 +42,15 @@ Leveraging a multi-role distributed architecture with Ray for flexible resource
 [Documents](https://alibaba.github.io/ROLL/)
 
 ### Quick Start
-[Installation](https://alibaba.github.io/ROLL/docs/English/QuickStart/installation)  
-[Config System Explanation](https://alibaba.github.io/ROLL/docs/English/QuickStart/config_system)  
-[Debugging Guide](https://alibaba.github.io/ROLL/docs/English/QuickStart/debugging_guide_en)  
-[Trackers and Metrics](https://alibaba.github.io/ROLL/docs/English/UserGuide/trackers_and_metrics)  
-[Checkpoint Saving and Resuming Guide](https://alibaba.github.io/ROLL/docs/English/UserGuide/checkpoint_and_resume)  
-[Converting MCoreAdapter Models to Hugging Face Format](https://alibaba.github.io/ROLL/docs/English/UserGuide/megatron_convert_2_hf)  
-[Quick Start: Single-Node Deployment Guide](https://alibaba.github.io/ROLL/docs/English/QuickStart/single_node_quick_start)  
-[Quick Start: Multi-Node Deployment Guide](https://alibaba.github.io/ROLL/docs/English/QuickStart/multi_node_quick_start)  
-[Frequently Asked Questions](https://alibaba.github.io/ROLL/docs/English/QuickStart/qa_issues)
+[Installation](https://alibaba.github.io/ROLL/docs/QuickStart/installation)  
+[Config System Explanation](https://alibaba.github.io/ROLL/docs/QuickStart/config_system)  
+[Debugging Guide](https://alibaba.github.io/ROLL/docs/QuickStart/debugging_guide_en)  
+[Trackers and Metrics](https://alibaba.github.io/ROLL/docs/UserGuide/trackers_and_metrics)  
+[Checkpoint Saving and Resuming Guide](https://alibaba.github.io/ROLL/docs/UserGuide/checkpoint_and_resume)  
+[Converting MCoreAdapter Models to Hugging Face Format](https://alibaba.github.io/ROLL/docs/UserGuide/megatron_convert_2_hf)  
+[Quick Start: Single-Node Deployment Guide](https://alibaba.github.io/ROLL/docs/QuickStart/single_node_quick_start)  
+[Quick Start: Multi-Node Deployment Guide](https://alibaba.github.io/ROLL/docs/QuickStart/multi_node_quick_start)  
+[Frequently Asked Questions](https://alibaba.github.io/ROLL/docs/QuickStart/qa_issues)
 
 ### UserGuide
 

diff --git a/...English/UserGuide/trackers_and_metrics.md → ...ll/docs/UserGuide/trackers_and_metrics.md b/...English/UserGuide/trackers_and_metrics.md → ...ll/docs/UserGuide/trackers_and_metrics.md
diff --git a/docs_roll/docs/index.md b/docs_roll/docs/index.md
diff --git a/docs_roll/docusaurus.config.js b/docs_roll/docusaurus.config.js
@@ -30,7 +30,17 @@ const config = {
   // to replace "en" with "zh-Hans".
   i18n: {
     defaultLocale: 'en',
-    locales: ['en'],
+    locales: ['en', 'zh-Hans'],
+    localeConfigs: {
+      en: {
+        label: 'English',
+        direction: 'ltr',
+      },
+      'zh-CN': {
+        label: '简体中文',
+        direction: 'ltr',
+      },
+    },
   },
 
   presets: [
@@ -115,11 +125,11 @@ const config = {
             items: [
               {
                 label: 'ROLL单机实践手册',
-                to: '/docs/简体中文/快速开始/single_node_quick_start_cn',
+                to: '/docs/QuickStart/single_node_quick_start',
               },
               {
                 label: '配置指南',
-                to: '/docs/简体中文/快速开始/config_guide_cn',
+                to: '/docs/QuickStart/config_guide',
               },
             ],
           },
Original file line number	Diff line number	Diff line change
Expand Up		@@ -298,4 +298,4 @@ rollout_dump_dir: odps://odps_project/tables/table_name/ds=${model_name}

		A: You can debug the trajectory generation process by configuring log levels and adding custom logs. Trajectory data will be completely saved in JSON format for easy analysis.

		For multi-round interaction local debugging, refer to the documentation: [Debug Guide](../../../简体中文/快速开始/debug_guide.md)
		For multi-round interaction local debugging, refer to the documentation: [Debug Guide](../../QuickStart/debug_guide.md)