From a18bd5994624456a9346fd65f8d30fd8039a4123 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=AE=A6=E6=99=93=E7=8E=B2?= <3174348550@qq.com> Date: Thu, 30 Oct 2025 09:27:17 +0800 Subject: [PATCH] modify links in 2.7.1 --- docs/mindstudio/docs/source_zh_cn/guide/large_model.md | 8 ++++---- .../deepseek_r1_671b_w8a8_dp4_tp4_ep4.md | 2 +- .../user_guide/supported_features/parallel/parallel.md | 4 ++-- .../user_guide/supported_features/parallel/parallel.md | 2 +- 4 files changed, 8 insertions(+), 8 deletions(-) diff --git a/docs/mindstudio/docs/source_zh_cn/guide/large_model.md b/docs/mindstudio/docs/source_zh_cn/guide/large_model.md index 8f4259bf93..c7fb712572 100644 --- a/docs/mindstudio/docs/source_zh_cn/guide/large_model.md +++ b/docs/mindstudio/docs/source_zh_cn/guide/large_model.md @@ -4,14 +4,14 @@ ## 基于MindSpore TransFormers大模型套件的调试调优指南 -[MindSpore TransFormers](https://www.mindspore.cn/mindformers/docs/zh-CN/master/index.html)是MindSpore提供的包含大模型预训练、微调、推理、部署的全流程开发套件，也是MindSpore当前常用的大模型开发套件。 +[MindSpore TransFormers](https://www.mindspore.cn/mindformers/docs/zh-CN/r1.7.0/index.html)是MindSpore提供的包含大模型预训练、微调、推理、部署的全流程开发套件，也是MindSpore当前常用的大模型开发套件。 -我们总结了大模型训练过程中常见精度问题、通用的精度问题定位方法、精度基准以及大模型场景工具常见用法，详见[大模型精度调试指南](https://www.mindspore.cn/mindformers/docs/zh-CN/master/advanced_development/precision_optimization.html#)。 +我们总结了大模型训练过程中常见精度问题、通用的精度问题定位方法、精度基准以及大模型场景工具常见用法，详见[大模型精度调试指南](https://www.mindspore.cn/mindformers/docs/zh-CN/r1.7.0/advanced_development/precision_optimization.html#)。 为了方便用户进行性能调优，MindSpore TransFormers套件集成了以下功能： -- 集成了Profiler数据采集的功能，并提供了超参可直接通过模型参数配置使用，详见[大模型性能调试指南](https://www.mindspore.cn/mindformers/docs/zh-CN/master/advanced_development/performance_optimization.html#)。 -- 集成了精度在线监控功能，详见[训练指标监控](https://www.mindspore.cn/mindformers/docs/zh-CN/master/feature/monitor.html)。 +- 集成了Profiler数据采集的功能，并提供了超参可直接通过模型参数配置使用，详见[大模型性能调试指南](https://www.mindspore.cn/mindformers/docs/zh-CN/r1.7.0/advanced_development/performance_optimization.html#)。 +- 集成了精度在线监控功能，详见[训练指标监控](https://www.mindspore.cn/mindformers/docs/zh-CN/r1.7.0/feature/monitor.html)。 ## 基于MindSpeed加速库的调试调优指南 diff --git a/docs/vllm_mindspore/docs/source_en/getting_started/tutorials/deepseek_parallel/deepseek_r1_671b_w8a8_dp4_tp4_ep4.md b/docs/vllm_mindspore/docs/source_en/getting_started/tutorials/deepseek_parallel/deepseek_r1_671b_w8a8_dp4_tp4_ep4.md index 0a786c659c..370bd1c350 100644 --- a/docs/vllm_mindspore/docs/source_en/getting_started/tutorials/deepseek_parallel/deepseek_r1_671b_w8a8_dp4_tp4_ep4.md +++ b/docs/vllm_mindspore/docs/source_en/getting_started/tutorials/deepseek_parallel/deepseek_r1_671b_w8a8_dp4_tp4_ep4.md @@ -153,7 +153,7 @@ Environment Variable Descriptions: - `MS_ENABLE_LCCL`: Disables LCCL and enables HCCL communication. - `HCCL_OP_EXPANSION_MODE`: Configures the scheduling and expansion location of the communication algorithm to be the AI Vector Core computing unit on the Device side. -- `MS_ALLOC_CONF`: Sets the memory policy. Refer to the [MindSpore Official Documentation](https://www.mindspore.cn/docs/en/master/api_python/env_var_list.html). +- `MS_ALLOC_CONF`: Sets the memory policy. Refer to the [MindSpore Official Documentation](https://www.mindspore.cn/docs/en/r2.7.1/api_python/env_var_list.html). - `ASCEND_RT_VISIBLE_DEVICES`: Configures the available device IDs for each node. Users can query this using the `npu-smi info` command. - `VLLM_MS_MODEL_BACKEND`: The backend of the model being run. The models and model backends currently supported by the vLLM-MindSpore plugin can be queried in the [Model Support List](../../../user_guide/supported_models/models_list/models_list.md). - `PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION`: Used when there are version compatibility issues. diff --git a/docs/vllm_mindspore/docs/source_en/user_guide/supported_features/parallel/parallel.md b/docs/vllm_mindspore/docs/source_en/user_guide/supported_features/parallel/parallel.md index 7d529f196c..a8879ab3bb 100644 --- a/docs/vllm_mindspore/docs/source_en/user_guide/supported_features/parallel/parallel.md +++ b/docs/vllm_mindspore/docs/source_en/user_guide/supported_features/parallel/parallel.md @@ -1,6 +1,6 @@ # Parallel Inference Methods -[![View Source](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/master/resource/_static/logo_source_en.svg)](https://gitee.com/mindspore/docs/blob/r2.7.1/docs/vllm_mindspore/docs/source_en/user_guide/supported_features/parallel/parallel.md) +[![View Source](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/r2.7.1/resource/_static/logo_source_en.svg)](https://gitee.com/mindspore/docs/blob/r2.7.1/docs/vllm_mindspore/docs/source_en/user_guide/supported_features/parallel/parallel.md) The vLLM-MindSpore plugin supports hybrid parallel inference configurations combining Tensor Parallelism (TP), Data Parallelism (DP), and Expert Parallelism (EP), and can be launched for multi-node multi-card setups using `Ray` or `multiprocess`. For applicable scenarios of different parallel strategies, refer to the [vLLM Official Documentation](https://docs.vllm.ai/en/latest/configuration/optimization.html#parallelism-strategies). The following sections will detail the usage scenarios, parameter configuration, and [Online Inference](#online-inference) for [Tensor Parallelism](#tensor-parallelism), [Data Parallelism](#data-parallelism), [Expert Parallelism](#expert-parallelism), and [Hybrid Parallelism](#hybrid-parallelism). @@ -237,7 +237,7 @@ Environment Variable Descriptions: - `MS_ENABLE_LCCL`: Disables LCCL and enables HCCL communication. - `HCCL_OP_EXPANSION_MODE`: Configures the scheduling and expansion location of the communication algorithm to be the AI Vector Core computing unit on the Device side. -- `MS_ALLOC_CONF`: Sets the memory policy. Refer to the [MindSpore Official Documentation](https://www.mindspore.cn/docs/en/master/api_python/env_var_list.html). +- `MS_ALLOC_CONF`: Sets the memory policy. Refer to the [MindSpore Official Documentation](https://www.mindspore.cn/docs/en/r2.7.1/api_python/env_var_list.html). - `ASCEND_RT_VISIBLE_DEVICES`: Configures the available device IDs for each node. Users can query this using the `npu-smi info` command. - `VLLM_MS_MODEL_BACKEND`: The backend of the model being run. The models and model backends currently supported by the vLLM-MindSpore plugin can be queried in the [Model Support List](../../../user_guide/supported_models/models_list/models_list.md). diff --git a/docs/vllm_mindspore/docs/source_zh_cn/user_guide/supported_features/parallel/parallel.md b/docs/vllm_mindspore/docs/source_zh_cn/user_guide/supported_features/parallel/parallel.md index 359cf45f81..51bc83936d 100644 --- a/docs/vllm_mindspore/docs/source_zh_cn/user_guide/supported_features/parallel/parallel.md +++ b/docs/vllm_mindspore/docs/source_zh_cn/user_guide/supported_features/parallel/parallel.md @@ -237,7 +237,7 @@ export VLLM_MS_MODEL_BACKEND=MindFormers - `MS_ENABLE_LCCL`：关闭LCCL，使能HCCL通信。 - `HCCL_OP_EXPANSION_MODE`：配置通信算法的编排展开位置为Device侧的AI Vector Core计算单元。 -- `MS_ALLOC_CONF`：设置内存策略。可参考[MindSpore官网文档](https://www.mindspore.cn/docs/zh-CN/master/api_python/env_var_list.html)。 +- `MS_ALLOC_CONF`：设置内存策略。可参考[MindSpore官网文档](https://www.mindspore.cn/docs/zh-CN/r2.7.1/api_python/env_var_list.html)。 - `ASCEND_RT_VISIBLE_DEVICES`：配置每个节点可用device id。用户可使用`npu-smi info`命令进行查询。 - `VLLM_MS_MODEL_BACKEND`：所运行的模型后端。目前vLLM-MindSpore插件所支持的模型与模型后端，可在[模型支持列表](../../../user_guide/supported_models/models_list/models_list.md)中进行查询。 -- Gitee