各位开发者们，本篇是为大家整理的实操教程集锦，包括官方输出、优秀开发者分享的实践案例。(持续更新）。请作为能量输入，开始构建专属的创意AI Agent吧~ <ModelArts Versatile-AI原生应用引擎 体验入口> 华为开发者空间 -- 开发平台 --Versatile Agent（ 请在PC端打开 ） ——官方指导案例案例标题入口展示平台超详细攻略：教你3分钟在华为开发者空间构建专属AgentLINK博客-开发者空间快速使用华为开发者空间AI Agent打造你的私人营养师LINK论坛-开发者空间基于华为开发者空间开发平台 MCP资产快速构建AI Agent应用 LINK案例中心基于华为开发者空间开发平台构建We码会议助手LINK案例中心基于开发者空间开发平台工作流构建旅行行程规划应用LINK案例中心华为云ModelArts Versatile训练营基础实验手册——零基础秒变大师！快速开发帮你打造爆款AI Agent：热点新闻助手LINK论坛-开天aPaaS华为云ModelArts Versatile训练营基础实验手册——零基础秒变大师！快速开发帮你打造爆款AI Agent：出行规划助手LINK论坛-开天aPaaS ——精选用户共创案例案例标题入口展示平台案例贡献用户【案例共创】基于华为开发者空间构建实时股票分析助手LINK案例中心miyalian【案例共创】基于华为开发者空间-AI Agent开发平台构建旅游规划助手LINK案例中心yd_272483742【案例共创】基于华为开发者空间开发平台 MCP资产快速构建税务AI助手服务LINK案例中心小草飞上天【案例共创】基于华为云开发者空间-Versatile Agent开发平台构建昇腾C算子开发知识库LINK案例中心黄生【案例共创】基于华为云开发者空间的AI Agent【旅行灵感生成器】智能体LINK案例中心柠檬味拥抱  查看更多华为云社区-案例中心入口：cid:link_12

openai 接口测试，耗时 58 秒响应日志  启动脚本 显存占用情况

如题目前用了两张，还是报错[INFO|2025-08-19 18:40:33] llamafactory.api.chat:143 >> ==== request ==== { "model": "Qwen3-Coder-30B-A3B-Instruct", "messages": [ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "hello" } ] } [WARNING|logging.py:328] 2025-08-19 18:40:34,022 >> `generation_config` default values have been modified to match model-specific defaults: {'top_k': 20, 'repetition_penalty': 1.05}. If this is not desired, please set these values explicitly. [E819 18:41:01.075574499 compiler_depend.ts:422] call aclnnIndexAdd failed, detail:EL0004: [PID: 39326] 2025-08-19-18:40:37.347.771 Failed to allocate memory. Possible Cause: Available memory is insufficient. Solution: Close applications not in use. TraceBack (most recent call last): Check param failed, mdl can not be null.[FUNC:LaunchKernelPrepare][FILE:context.cc][LINE:901] kernel launch prepare failed.[FUNC:LaunchKernelWithHandle][FILE:context.cc][LINE:1325] rtKernelLaunchWithHandleV2 execute failed, reason=[module new memory error][FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:53] rtKernelLaunchWithHandleV2 failed: 207001 #### KernelLaunch failed: /usr/local/Ascend/ascend-toolkit/8.1.RC1/opp/built-in/op_impl/ai_core/tbe//kernel/ascend910b/slice/Slice_a32d70ed63082227463347a44bd5a08e_high_performance.o Kernel Run failed. opType: 43, Slice launch failed for Slice, errno:361001. Task execute failed, device_id=1, stream_id=4, task_id=0, flip_num=0, task_type=87.[FUNC:GetError][FILE:stream.cc][LINE:1119] Failed to synchronize stream, retCode=0x7150004.[FUNC:SyncGetDevMsg][FILE:api_impl.cc][LINE:5918] Sync get device msg failed, retCode=0x7150004.[FUNC:GetDevErrMsg][FILE:api_impl.cc][LINE:5937] rtGetDevMsg execute failed, reason=[tsfw param illegal][FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:53] [ERROR] 2025-08-19-18:41:01 (PID:39326, Device:1, RankID:-1) ERR01100 OPS call acl api failed Exception raised from operator() at build/CMakeFiles/torch_npu.dir/compiler_depend.ts:37 (most recent call first): frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0xb8 (0xffff9460f908 in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/lib/libc10.so) frame #1: c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::string const&) + 0x6c (0xffff945be404 in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/lib/libc10.so) frame #2: <unknown function> + 0xe90bd8 (0xfffdec492bd8 in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch_npu/lib/libtorch_npu.so) frame #3: <unknown function> + 0x1644484 (0xfffdecc46484 in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch_npu/lib/libtorch_npu.so) frame #4: <unknown function> + 0x78d244 (0xfffdebd8f244 in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch_npu/lib/libtorch_npu.so) frame #5: <unknown function> + 0x78da58 (0xfffdebd8fa58 in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch_npu/lib/libtorch_npu.so) frame #6: <unknown function> + 0x78a1cc (0xfffdebd8c1cc in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch_npu/lib/libtorch_npu.so) frame #7: <unknown function> + 0x4c9e4c (0xffff9464ce4c in /home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/lib/libc10.so) frame #8: <unknown function> + 0x87c4 (0xffffabd3f7c4 in /usr/lib64/libpthread.so.0) frame #9: <unknown function> + 0xdbcec (0xffffabb79cec in /usr/lib64/libc.so.6) INFO: 127.0.0.1:41030 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error ERROR: Exception in ASGI application Traceback (most recent call last): File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/uvicorn/protocols/http/h11_impl.py", line 403, in run_asgi result = await app( # type: ignore[func-returns-value] File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/uvicorn/middleware/proxy_headers.py", line 60, in __call__ return await self.app(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/fastapi/applications.py", line 1054, in __call__ await super().__call__(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/applications.py", line 113, in __call__ await self.middleware_stack(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/errors.py", line 186, in __call__ raise exc File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/errors.py", line 164, in __call__ await self.app(scope, receive, _send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/cors.py", line 85, in __call__ await self.app(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/exceptions.py", line 63, in __call__ await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app raise exc File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app await app(scope, receive, sender) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 716, in __call__ await self.middleware_stack(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 736, in app await route.handle(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 290, in handle await self.app(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 78, in app await wrap_app_handling_exceptions(app, request)(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app raise exc File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app await app(scope, receive, sender) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 75, in app response = await f(request) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/fastapi/routing.py", line 302, in app raw_response = await run_endpoint_function( File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/fastapi/routing.py", line 213, in run_endpoint_function return await dependant.call(**values) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/api/app.py", line 110, in create_chat_completion return await create_chat_completion_response(request, chat_model) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/api/chat.py", line 189, in create_chat_completion_response responses = await chat_model.achat( File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/chat/chat_model.py", line 92, in achat return await self.engine.chat(messages, system, tools, images, videos, audios, **input_kwargs) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/chat/hf_engine.py", line 363, in chat return await asyncio.to_thread(self._chat, *input_args) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/asyncio/threads.py", line 25, in to_thread return await loop.run_in_executor(None, func_call) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/concurrent/futures/thread.py", line 52, in run result = self.fn(*self.args, **self.kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/chat/hf_engine.py", line 240, in _chat generate_output = model.generate(**gen_kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/generation/utils.py", line 2460, in generate result = self._sample( File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/generation/utils.py", line 3426, in _sample outputs = self(**model_inputs, return_dict=True) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/accelerate/hooks.py", line 175, in new_forward output = module._old_forward(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/utils/generic.py", line 965, in wrapper output = func(self, *args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/utils/deprecation.py", line 172, in wrapped_func return func(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen3_moe/modeling_qwen3_moe.py", line 1043, in forward outputs: MoeModelOutputWithPast = self.model( File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/utils/generic.py", line 965, in wrapper output = func(self, *args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen3_moe/modeling_qwen3_moe.py", line 673, in forward layer_outputs = decoder_layer( File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/accelerate/hooks.py", line 175, in new_forward output = module._old_forward(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen3_moe/modeling_qwen3_moe.py", line 391, in forward hidden_states = self.mlp(hidden_states) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/accelerate/hooks.py", line 175, in new_forward output = module._old_forward(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen3_moe/modeling_qwen3_moe.py", line 277, in forward idx, top_x = torch.where(expert_mask[expert_idx]) RuntimeError: The Inner error is reported as above. The process exits for this inner error, and the current working operator name is aclnnIndexAdd. Since the operator is called asynchronously, the stacktrace may be inaccurate. If you want to get the accurate stacktrace, pleace set the environment variable ASCEND_LAUNCH_BLOCKING=1. Note: ASCEND_LAUNCH_BLOCKING=1 will force ops to run in synchronous mode, resulting in performance degradation. Please unset ASCEND_LAUNCH_BLOCKING in time after debugging. [ERROR] 2025-08-19-18:41:25 (PID:39326, Device:1, RankID:-1) ERR00100 PTA call acl api failed. EL0004: [PID: 39326] 2025-08-19-18:41:01.882.925 Failed to allocate memory. Possible Cause: Available memory is insufficient. Solution: Close applications not in use. TraceBack (most recent call last): Task execute failed, device_id=1, stream_id=5, task_id=0, flip_num=0, task_type=87.[FUNC:GetError][FILE:stream.cc][LINE:1119] Failed to synchronize stream, retCode=0x7150004.[FUNC:SyncGetDevMsg][FILE:api_impl.cc][LINE:5918] Sync get device msg failed, retCode=0x7150004.[FUNC:GetDevErrMsg][FILE:api_impl.cc][LINE:5937] rtGetDevMsg execute failed, reason=[tsfw param illegal][FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:53] EL0004: [PID: 39326] 2025-08-19-18:40:37.347.771 Failed to allocate memory. Possible Cause: Available memory is insufficient. Solution: Close applications not in use. TraceBack (most recent call last): Check param failed, mdl can not be null.[FUNC:LaunchKernelPrepare][FILE:context.cc][LINE:901] kernel launch prepare failed.[FUNC:LaunchKernelWithHandle][FILE:context.cc][LINE:1325] rtKernelLaunchWithHandleV2 execute failed, reason=[module new memory error][FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:53] rtKernelLaunchWithHandleV2 failed: 207001 #### KernelLaunch failed: /usr/local/Ascend/ascend-toolkit/8.1.RC1/opp/built-in/op_impl/ai_core/tbe//kernel/ascend910b/slice/Slice_a32d70ed63082227463347a44bd5a08e_high_performance.o Kernel Run failed. opType: 43, Slice launch failed for Slice, errno:361001. Task execute failed, device_id=1, stream_id=4, task_id=0, flip_num=0, task_type=87.[FUNC:GetError][FILE:stream.cc][LINE:1119] Failed to synchronize stream, retCode=0x7150004.[FUNC:SyncGetDevMsg][FILE:api_impl.cc][LINE:5918] Sync get device msg failed, retCode=0x7150004.[FUNC:GetDevErrMsg][FILE:api_impl.cc][LINE:5937] rtGetDevMsg execute failed, reason=[tsfw param illegal][FUNC:FuncErrorReason][FILE:error_message_manage.cc][LINE:53]

训练模型用的大规模数据集一般怎么存储呀？是直接用 云硬盘EVS，还是用 对象存储服务OBS？

<ModelArts Versatile-AI原生应用引擎 体验入口> 华为开发者空间 -- 开发平台 --Versatile Agent（ 请在PC端打开 ）  什么是AI AgentAI Agent人工智能体，是一种能够自主感知环境、制定目标、规划行动、执行任务并持续学习的智能程序或系统。比如，告诉AI Agent帮忙下单一份外卖，它就可以直接调用 APP选择外卖，再调用支付程序下单支付，无需人类去指定每一步的操作。从本质上来讲，大模型作为大脑发号施令，推理能力决定它的决策能力上限，工具模块决定他行动能力的上限、可完成任务的横向宽度、操作场域。· AI Agent之ModelArts VersatileModelArts Versatile-AI原生应用引擎是一站式企业级全生命周期的智能体平台，为企业专属大模型应用开发的工具链，提供灵活的画布式AI Agent开发能力，让Agent能够准确解决复杂的业务场景问题。提供从Agent开发，使用，运营和运维的全生命周期管理能力。  快速了解 AI Agent的插件——MCP/工具 · MCP2024年11月底 Anthropic（Claude提供商）提出了MCP协议，并在Claude 客户端支持了MCP。MCP，全称Model Context Protocol，模型上下文协议。被认为是未来AI生态的「标准USB接口」，AI界的 “万能转换器”。MCP 是一个开放协议，核心目标是打通大语言模型（LLM）应用与外部数据源、工具之间的交互壁垒。由于每项数据源、工具或服务都有独立的格式规范、对接协议和认证体系，在传统的Agent开发场景中，如果AI模型需要与各类外部资源对接，开发者需要针对每个API进行独立的代码、文档、认证方式、错误处理和后续维护，这意味着每一种具体的连接都需要编写代码，效率低下且耗时费力。而MCP的出现，如同在AI模型与外部世界之间搭建了一座标准化的桥梁。MCP以通用的“标准语言”把工具、数据通过“MCP服务器”的方式供给（一次开发、无限连接），可以更高效、更便捷地实现Agent与成千上万的外部工具与数据的互通。 Before MCP：每个AI应用都需要单独封装API对接相同的外部系统，存在碎片化集成的问题。With MCP：统一抽象AI应用与外部系统的交互的标准化接口，不同AI应用通过相同Client即可连接到对应的Server。 MCP核心组件MCP Host：LLM应用程序（例如Claude Desktop，Agent智能体应用）MCP Client（MCP客户端）：嵌入在MCP Host中，MCP Client对象与MCP Server的关系是1:1MCP Server（MCP服务器）：向MCP Client提供上下文（例如外置的对话历史），工具执行，prompt模板等能力MCP Protocol（MCP协议）：MCP Client和MCP Server之间数据传输的协议，当前支持本地stdio和远程http sse两种模式 其中MCP Server能力-工具执行：向AI应用提供工具执行能力，例如网页搜索、发送消息、执行LLM生成的代码等-资源访问：向AI应用提供资源访问的能力，例如访问文件系统、数据库等资源-Prompt模板：向AI应用提供预定义好的Prompt或工作流模板，指导LLM输出更准确的内容· 工具工具是一组相关的API集合，一个工具通常包含多个执行动作，每个执行动作用于实现特定功能，工具可以是函数、算法、或者是其他任何能够帮助Agent提高工作效率的辅助性资源。在创建Agent时调用工具，可以有效提高Agent开发的效率。例如，在人工智能领域，智能体可以通过调用不同的函数或算法来实现特定的功能，这些函数或算法可以被视为智能体的工具。通过这些工具的支持，智能体能够更加灵活地应对各种复杂任务和环境变化，从而提高其解决问题的效率和准确性。在创建工具时，需要先将选定的API服务注册为一个工具，然后再添加该服务下的API作为工具的执行动作。 ModelArts Versatile--AI原生应用引擎 MCP资产与工具的竞争力优势 · MCP资产 技术亮点01 ModelArts Versatile平台预置海量MCP资产，一键安装，开箱即用无码化开发，无需额外部署，预置MCP资产可实现Versatile AI Agent（智能体）能力边界拓展延伸。 02 兼容主流的三方插件及MCP接入标准，无缝集成MCP通过统一协议替代传统"MXN"式开发模式，使AI应用可通过单一接口访问文件、数据库、API、第三方工具等异构资源，降低开发复杂度300%以上。同时具备动态兼容性：当新增工具时，只需将其注册到MCP Server，所有接入MCP的LLM均可立即调用，无需重新训练或调整模型。 03 基于MoA架构和MCP协议，实现生态扩展及Multi Agent架构服务提供商按照MCP协议发布服务，一次发布。模型提供商按照MCP协议集成服务，一次集成。 相关特性ModelArts Versatile资产中心提供多种MCP资源，用户通过简单安装即可快速集成调用；平台支持灵活拓展，兼容开源社区MCP及自主开发MCP服务的接入。   · 工具 技术亮点01 预置60+常用工具资产Tools，让Agent天生具备强大问题解决能力，开箱即用ModelArts Versatile平台在资产中心预置了丰富的工具，拓展能力边界，突破模型固有局限，连接真实世界需求，提供广泛能力，并实现复杂任务自动化。 02 自定义插件工具，极大丰富智能体功能生态Versatile平台提供插件定制化体验、帮助用户构建专属工具、从广场选择他人插件/智能添加。同时也支持用户创建、上架、编辑、导入更新等工具管理方式，精准满足特定业务场景的需求。 03 灵活配置调用工具，保障Agent行动效率Agent支持工具识别、检索、调用；支持手动选择工具，支持工具搜索；支持根据基础信息自动选择工具； 相关特性资产库-预置工具：ModelArts Versatile资产中心展示了平台预置的第三方工具，这些工具可在创建Agent时便捷调用，同时可对60+工具可设置鉴权、收藏。创建工具：将选定的API服务注册为一个工具；添加API服务下的具体接口作为工具的执行动作。导入工具：平台支持通过导入OpenAPI规范文件（.json格式）自动解析并生成工具配置，从而提升工具创建效率。  · Versatile MCP资产与工具的差异与协作差异-工具：是AI 可调用的功能模块集合，实现智能体能力扩展。工具暂不适配MCP协议，部分为用户本地部署仅供内部使用，补齐能力。-MCP资产：该类资产适配当下大火的技术点——MCP协议，通过标准化协议实现AI模型与外部工具、数据源的高效连接。可在平台内一键安装，快速调用。而MCP标准化促进开发者生态的可持续快速发展。 协作单Agent (自主规划模式)：MCP+工具相辅相成，加速能力延伸，快速构建出千行百业的AI智能体。MCP资产是工具Tools类插件在技术迭代背景下的升级，而MCP为基于大模型的助手与代理系统提供一个通用的接口标准，可实现即刻连接海量外部工具。两者加持赋能ModelArts Versatile，面向开发者、伙伴、生态，加速推动AI原生应用领域生态繁荣，实现智能体价值闭环。  Versatile--AI原生应用引擎 MCP/工具主要解决什么问题（Agent编排中心） · MCP使 AI 能够获取实时、准确的上下文信息；让不同平台、服务之间无缝协作；MCP是 Function Calling （函数调用）和 Tools 高效运行的基础。 打破数据孤岛：传统大模型无法直接访问实时数据或本地资源，而MCP让AI“连接万物”，例如，查询天气时自动调用气象API，分析企业数据时直接连接内部数据库。降低开发成本：在MCP出现之前，每个大模型需要为每个工具单独开发接口，导致重复劳动。而通过MCP，开发者只需写一次服务端，所有兼容MCP的模型都能调用。提升安全性与互操作性：MCP内置权限控制和加密机制，比直接开放数据库更安全；同时，类似USB接口的标准化让不同厂商的工具能“即插即用”，避免生态分裂。 · 工具 ‌能力扩展‌：通过集成各种工具，AI Agent可以访问外部数据源、执行特定计算或操作，从而扩展其原生能力范围。‌任务执行‌：工具使AI Agent能将抽象决策转化为具体操作，如调用API、操作数据库、生成代码等。‌环境交互‌：通过传感器接口、设备控制等工具，AI Agent能感知物理/数字环境并施加影响。工业领域的IoT监控Agent就是通过专用工具实现设备状态采集。‌智能增强‌：特定工具可以提升Agent的推理、规划能力。如LangGraph框架中的状态机工具支持复杂业务流程的编排与回溯。  总体来说，ModelArts Versatile-AI原生应用引擎拥有丰富的插件生态，包含平台预置的海量MCP资产/工具、由用户自定义配置的实用插件等，达成Agent能力边界无限拓展。开发者们通过可视化轻松编排Agent，无码化分钟级构建智能体，推动AI生产方式革新，助力千行万业实现产业升级、数字化转型迈向新高度。    点击可前往>>华为云ModelArts Versatile-AI原生应用 引擎官网  

我是已收到代金券的老师，需要分配IAM子账户给50位学生使用modelarts，使用代金券；

请问关于下列不同款的ModelArts昇腾芯片的区别？ModelArts昇腾310(鲲鹏920)ModelArts昇腾训练NPU(鲲鹏910)ModelArts昇腾AI加速型(B1)ModelArts昇腾AI加速型(B2)ModelArts昇腾AI加速型(B3)这里：https://bbs.huaweicloud.com/forum/thread-0257165231016526001-1-1.html 有相关问题，但是跳转的帖子不存在了。

华为云ModelArts Versatile训练营基础实验之AI原生应用开发—零基础秒变大师！快速开发帮你打造爆款AI Agent：出行规划助手  一 基本信息实验类型实验难度：简单实验时长：30分钟实验类型：实操型实验简介本次实验将指导开发者通过零码快速构建复杂任务规划专家AI Agent“出行规划助手”，并在应用生成后体验试用，体验快速创建AI原生应用，轻松完成行业场景的乐趣实验目标完成出行Agent的从0创建自定义出行Agent描述，技能等与出行Agent交互，助手回答用户相关提问实验任务3.配置MCP服务信息4.创建并配置Agent信息2.创建MCP服务1.登录Versatile6.根据Agent返回确认结果5.点击开始体验登录Versatile界面点击创建MCP服务，进入MCP服务创建界面配置MCP服务信息创建并配置Agent信息点击开始体验，并在对话框中输入和出行Agent交互的内容根据Agent返回确认结果：根据出行Agent的返回内容，确认是否完成了用户的指令学前建议1. 了解相关技能2. 了解MCP服务3. 了解Agent知识二 实验实操步骤一：进入Versatile首页，左边菜单栏选择“我的MCP”点击创建MCP服务步骤二：配置MCP服务信息，部署12306和高德MCP服务1. 选择12306-MCP服务模板后点击下一步2. 点击安装MCP点击安装后，等待MCP服务安装完成，显示安装中。3. 选择高德地图MCP模板后点击下一步4. 配置高德api-key后点击安装MCP请填写api-key：API Key创建方式请见：cid:link_05. 等待MCP安装完成步骤三：创建并配置Agent信息1. 左边菜单栏选择“我的Agent”点击创建Agent。2. 选择单Agent(复杂任务规划)，配置Agent信息并添加前面部署的MCP服务，点击发布。经验模板填写：# 工具使用你可以使用12306相关mcp工具查询火车票相关信息。你可以使用高德相关mcp工具查询导航相关信息。3. 填写api-key，点击发布。请填写api-key：API Key创建方式请见：cid:link_0步骤四：旅游出行助手交互体验1. 在agent列表中，选择“旅游出行助手”体验2. 在对话框中输入：6月23日中午坐高铁从深圳到上海，请规划详细的行程路线规划，我早上从深圳华为基地出发，坐地铁到高铁站。最终用网页为我呈现，并等待Agent返回：生成计划后，点击开始任务，也可以进一步修改任务：最终生成网页进行浏览：3. 完成后，可自由提问相关问题。

华为云ModelArts Versatile训练营基础实验之AI原生应用开发—零基础秒变大师！快速开发帮你打造爆款AI Agent：热点新闻助手  一 基本信息实验类型实验难度：简单实验时长：30分钟实验类型：实操型实验简介本次实验将指导开发者通过零码快速构建复杂任务规划专家AI Agent“热点新闻助手”，并在应用生成后体验试用，体验快速创建AI原生应用，轻松完成行业场景的乐趣实验目标完成出行Agent的从0创建自定义出行Agent描述，技能等与出行Agent交互，助手回答用户相关提问实验任务3.配置MCP服务信息4.创建并配置Agent信息2.创建MCP服务1.登录Versatile6.根据Agent返回确认结果5.点击开始体验登录Versatile界面点击创建MCP服务，进入MCP服务创建界面配置MCP服务信息创建并配置Agent信息点击开始体验，并在对话框中输入和出行Agent交互的内容根据Agent返回确认结果：根据出行Agent的返回内容，确认是否完成了用户的指令学前建议1. 了解相关技能2. 了解MCP服务3. 了解Agent知识二 实验实操步骤一：进入Versatile首页，左边菜单栏选择“我的MCP”点击创建MCP服务步骤二：配置MCP服务信息，部署中文趋势聚合服务1. 选择“中文趋势聚合”服务模板后点击下一步2. 点击安装MCP3. 点击安装后，等待MCP服务安装完成，显示安装中步骤三：创建并配置Agent信息1. 左边菜单栏选择“我的Agent”点击创建Agent。2. 选择单Agent(复杂任务规划)，配置Agent信息并添加前面部署的MCP服务，点击发布。经验模板填写：# 工具使用- get-36kr-trending获取 36 氪热榜，提供创业、商业、科技领域的热门资讯，包含投融资动态、新兴产业分析和商业模式创新信息- get-bilibili-rank获取哔哩哔哩视频排行榜，包含全站、动画、音乐、游戏等多个分区的热门视频，反映当下年轻人的内容消费趋势- get-douban-rank获取豆瓣实时热门榜单，提供当前热门的图书、电影、电视剧、综艺等作品信息，包含评分和热度数据- get-douyin-trending获取抖音热搜榜单，展示当下最热门的社会话题、娱乐事件、网络热点和流行趋势- get-ifanr-news获取爱范儿科技快讯，包含最新的科技产品、数码设备、互联网动态等前沿科技资讯- get-netease-news-trending获取网易新闻热点榜，包含时政要闻、社会事件、财经资讯、科技动态及娱乐体育的全方位中文新闻资讯- get-tencent-news-trending获取腾讯新闻热点榜，包含国内外时事、社会热点、财经资讯、娱乐动态及体育赛事的综合性中文新闻资讯- get-toutiao-trending获取今日头条热榜，包含时政要闻、社会事件、国际新闻、科技发展及娱乐八卦等多领域的热门中文资讯3. 填写api-key，点击发布。请填写api-key：API Key创建方式请见：cid:link_0步骤四：热点新闻助手交互体验1. 在agent列表中，选择“热点新闻助手”体验2. 在对话框中输入：最近有哪些热点新闻、热门电影和即将上映的热门电影，请详细列举下，并用图文并茂的网页为我呈现，并等待Agent返回：生成计划后，点击开始任务，也可以进一步修改任务：最终生成网页进行浏览：3. 完成后，可自由提问相关问题。

问题1：如何在notebook上安装docker环境？(base) [ma-user infini_infer] $ which docker which: no docker in 问题2：机器的CA证书有问题(base) [ma-user infini_infer]$curl -fsSL https://xmake.io/shget.text | bash curl: (77) error setting certificate verify locations: CAfile: /etc/pki/tls/certs/ca-bundle.crt CApath: none如何提升至root来修复CA证书问题？

我想本地部署deepseek，借助modelarts平台昇腾算力，在要拉取镜像时发现，modelarts的管理员权限不开放，导致我无法下载docker，卡在这里，我想在这里求助一下，还有其他下载docker的方法吗？

      2025年8月8日，华为云MaaS大模型即服务平台实践技术沙龙在九云科教集团成功举办。本次活动由华为云HCDG长沙核心组周毅、黄蓉发起，来自于北京并行科技、北京惠农通正、杭州泰师科技、湖南中科博远、湖南拓维信息、湖南科创信息等企业的30余位人工智能领域开发者齐聚一堂，活动围绕当前热门的MaaS平台展开深入讨论，共谋AI大模型的应用与发展。       长沙九云信息科技有限公司董事长高宏祥发表致辞，高度赞赏华为作为中国科技领军企业的担当精神，特别提到其自主创新的鸿蒙系统、昇腾AI等突破性成果。他希望与会开发者把握AI时代机遇，依托华为云等开放平台，共同探索智能教育、行业大模型等创新应用，推动AI生态高质量发展。       华为云DTSE技术专家分享了华为云在AI大模型和数字化转型的成功实践和技术积累，介绍了如何使用ModelArts Studio一站式大模型开发平台完成模型的训推与部署，结合平台提供的模型API、MCP、应用模板快速构建应用，实现行业应用的高效落地，并与企业开发者们深度探讨了平台底层能力。       华为云DTSE高级工程师在活动中，带领大家在华为开发者空间中体验了《基于DeepSeek和Dify构建心理咨询师应用》的案例，此案例使用Dify对接华为云MaaS平台，成功连接到基于昇腾训练的DeepSeek大模型，实现了心理咨询师应用的功能。这个案例的实操，让开发者了解到华为开发者空间为开发者提供的高效、稳定的开发环境，同时华为云MaaS平台作为Dify内置模型供应商，对接更加便捷。       华为云战略合作伙伴——金蝶软件的院校高级顾问杨名先生分享了“智联云途：华为云携手金蝶云苍穹探索智能时代新路径”主题演讲。金蝶云苍穹ERP深度融合华为云AI平台，通过技术共生、场景共创、生态共享，正在重塑企业数字化转型的路径。从底层架构的自主可控，到行业场景的深度渗透，再到开发者生态的繁荣，双方不仅为企业提供 “开箱即用” 的智能解决方案，更通过持续的技术创新与模式探索，引领智能时代的产业变革。这种 “云+端+AI”的协同模式，将成为推动中国企业数智化升级的核心引擎。       华为云伙伴——网久软件联合创始人&COO周建军先生分享了“基于华为云部署开源组件的方法”主题演讲，Websoft9 多应用托管与运维平台可以通过华为云商店一键购买并创建ECS实例，自动完成部署。支持多应用托管、可视化运维及资源扩展，10分钟内即可上线。       在集中讨论环节中，与会企业围绕MaaS平台、Agent开发技术、AI人才培养等展开深度交流。大家一致认为，MaaS平台通过提供预训练模型和低代码工具，可显著降低企业AI应用开发门槛；2025年被视为Agent技术商业化落地的关键元年，各行业都在积极探索智能体在业务流程优化、客户服务等场景的应用价值；而针对AI人才培养问题，与会代表建议通过校企合作建立实训基地，并开发体系化培训课程，重点培养既懂AI技术又熟悉行业应用的复合型人才，以支撑企业智能化转型需求。       华为云高级技术专家毛定宇为九云科教集团的产品总监周毅和运维部经理黄蓉，授予华为云开发者组织HCDG长沙站旗帜。这对于中部地区的开发人员来说是一个重要里程碑时刻，再次感谢长沙HCDG核心组成员们对于本次活动的辛勤付出和认真负责，相信长沙HCDG会在以后举办出更多有价值有意义的精彩活动、赋能越来越多的开发者。       HCDG（Huawei Cloud Developer Group 华为云开发者社区组织），是基于城市圈和技术圈，由开发者核心组自发开展的开放、创新、多元的社区技术交流组织。致力于帮助开发者学习提升、互动交流、挖掘合作，推动技术应用与本地产业结合、数智化转型和开发者文化发展。迄今为止，越来越多的社区和组织开始广泛的与HCDG进行密切的合作与往来，相信在今后的日子里，HCDG会以更加崭新的面貌出现在开发者视野中，我们期待越来越多优秀的开发者、创业者、运营人、产品人、老师、学生等加入到HCDG，让HCDG茁壮成长、遍地开花、节节高！

镜像：pytorch_2.1.0-cuda_12.1-py_3.10.6-ubuntu_22.04-x86_64pip list：Package Version Editable project location ---------------------- ----------- -------------------------------- absl-py 2.3.1 accelerate 1.7.0 aiofiles 24.1.0 aiohappyeyeballs 2.6.1 aiohttp 3.12.15 aiosignal 1.4.0 annotated-types 0.7.0 antlr4-python3-runtime 4.9.3 anyio 4.10.0 async-timeout 5.0.1 attrs 25.3.0 audioread 3.0.1 auto_tune 0.1.0 av 15.0.0 certifi 2025.8.3 cffi 1.17.1 chardet 5.2.0 charset-normalizer 3.4.2 click 8.2.2 colorama 0.4.6 contourpy 1.3.2 cycler 0.12.1 dataflow 0.0.1 DataProperty 1.1.0 datasets 2.21.0 decorator 5.2.1 deepspeed 0.15.4 diffusers 0.31.0 dill 0.3.8 docker 7.1.0 docstring_parser 0.17.0 einops 0.8.0 evaluate 0.4.1 exceptiongroup 1.3.0 fastapi 0.116.1 ffmpy 0.6.1 filelock 3.18.0 fire 0.7.0 fonttools 4.59.0 frozenlist 1.7.0 fsspec 2024.6.1 gradio 5.31.0 gradio_client 1.10.1 groovy 0.1.2 h11 0.16.0 hccl 0.1.0 hccl_parser 0.1 hf_transfer 0.1.9 hf-xet 1.1.5 hjson 3.1.0 httpcore 1.0.9 httpx 0.28.1 huggingface-hub 0.34.3 idna 3.10 importlib_metadata 8.7.0 jieba 0.42.1 Jinja2 3.1.6 joblib 1.5.1 jsonlines 4.0.0 kiwisolver 1.4.8 lazy_loader 0.4 librosa 0.11.0 llamafactory 0.9.4.dev0 /home/ma-user/work/LLaMA-Factory llm_datadist 0.0.1 llvmlite 0.44.0 lm_eval 0.4.3 lxml 6.0.0 markdown-it-py 3.0.0 MarkupSafe 3.0.2 matplotlib 3.10.5 mbstrdecoder 1.1.4 mdurl 0.1.2 modelscope 1.28.1 more-itertools 10.7.0 mpmath 1.3.0 msgpack 1.1.1 msobjdump 0.1.0 multidict 6.6.3 multiprocess 0.70.16 networkx 3.4.2 ninja 1.11.1.4 nltk 3.9.1 numba 0.61.2 numexpr 2.11.0 numpy 1.23.5 omegaconf 2.3.0 op_compile_tool 0.1.0 op_gen 0.1 op_test_frame 0.1 opc_tool 0.1.0 openmind 1.0.0 openmind-hub 0.9.1 orjson 3.11.1 packaging 25.0 pandas 2.3.1 pathvalidate 3.3.1 peft 0.14.0 pillow 11.3.0 pip 25.1 platformdirs 4.3.8 pooch 1.8.2 portalocker 3.2.0 propcache 0.3.2 protobuf 6.31.1 psutil 7.0.0 py-cpuinfo 9.0.0 pyarrow 21.0.0 pybind11 3.0.0 pycparser 2.22 pydantic 2.10.6 pydantic_core 2.27.2 pydub 0.25.1 Pygments 2.19.2 pyparsing 3.2.3 pytablewriter 1.2.1 python-dateutil 2.9.0.post0 python-multipart 0.0.20 pytz 2025.2 PyYAML 6.0.2 qwen-vl-utils 0.0.11 regex 2025.7.34 requests 2.32.2 responses 0.18.0 rich 14.1.0 rouge-chinese 1.0.3 rouge_score 0.1.2 ruff 0.12.7 sacrebleu 2.5.1 safehttpx 0.1.6 safetensors 0.6.1 schedule_search 0.0.1 scikit-learn 1.7.1 scipy 1.15.3 semantic-version 2.10.0 sentencepiece 0.2.0 setuptools 69.5.1 shellingham 1.5.4 show_kernel_debug_data 0.1.0 shtab 1.7.2 six 1.17.0 sniffio 1.3.1 soundfile 0.13.1 soxr 0.5.0.post1 sqlitedict 2.1.0 sse-starlette 3.0.2 starlette 0.47.2 sympy 1.13.1 tabledata 1.3.4 tabulate 0.9.0 tcolorpy 0.1.7 te 0.4.0 termcolor 3.1.0 threadpoolctl 3.6.0 tiktoken 0.10.0 tokenizers 0.21.1 tomlkit 0.13.3 torch 2.5.1 torch-npu 2.5.1 torchvision 0.20.1 tqdm 4.67.1 tqdm-multiprocess 0.0.11 transformers 4.51.0 trl 0.9.6 typepy 1.3.4 typer 0.16.0 typing_extensions 4.14.1 typing-inspection 0.4.1 tyro 0.8.14 tzdata 2025.2 urllib3 2.5.0 uvicorn 0.35.0 websockets 15.0.1 wheel 0.45.1 word2number 1.1 xxhash 3.5.0 yarl 1.20.1 zipp 3.23.0 zstandard 0.23.0 启动脚本$cat start-qwen2.5-vl-7b-instruct.sh #!/bin/bash source /usr/local/Ascend/ascend-toolkit/set_env.sh conda activate llamafactory export API_MODEL_NAME=Qwen2.5-VL-7B-Instruct export ASCEND_RT_VISIBLE_DEVICES=0 export API_PORT=18000 export model_path=/home/ma-user/work/model/${API_MODEL_NAME} export ASCEND_LAUNCH_BLOCKING=1 export FLASH_ATTENTION=enable export FRAMEWORK_VERSION=6.5.905+ export ASCEND_SLOG_PRINT_TO_STDOUT=1 nohup llamafactory-cli api \ --model_name_or_path ${model_path} \ --template qwen2_vl \ --infer_backend huggingface \ --trust_remote_code > ${API_MODEL_NAME}.log 2>&1 & 调用model=Qwen2.5-VL-7B-Instruct curl -X POST "http://${url}/v1/chat/completions" \ -H "Content-Type: application/json" \ -d '{ "model": "'"${model}"'", "messages": [ {"role":"system","content":[{"type": "text", "text": "You are a helpful assistant."}]}, {"role": "user", "content": [ {"type": "image_url", "image_url": {"url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg"}}, {"type": "text", "text": "图中描绘的是什么景象?"} ] }] }'报错[INFO|2025-08-07 19:39:46] llamafactory.api.chat:143 >> ==== request ==== { "model": "Qwen2.5-VL-7B-Instruct", "messages": [ { "role": "system", "content": [ { "type": "text", "text": "You are a helpful assistant." } ] }, { "role": "user", "content": [ { "type": "image_url", "image_url": { "url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg" } }, { "type": "text", "text": "图中描绘的是什么景象?" } ] } ] } [INFO] RUNTIME(2162556,python3.1):2025-08-07-19:39:46.531.061 [api_impl.cc:7464] 2165043 PeekLastErr: level=0 err=0. ... ... [ERROR] OP(2162556,python3.1):2025-08-07-19:39:55.937.196 [flash_attention_score_tiling_general.cpp:1496][OP_TILING][AnalyzeOptionalInput][2165043] OpName:[FlashAttentionScore:FlashAttentionScore] "atten mask dim should be 2 or 4, but got 3" [ERROR] OP(2162556,python3.1):2025-08-07-19:39:55.937.397 [flash_attention_score_tiling_general.cpp:847][OP_TILING][GetShapeAttrsInfo][2165043] OpName:[FlashAttentionScore] "fail to analyze context info." [ERROR] OP(2162556,python3.1):2025-08-07-19:39:55.937.458 [kernel_workspace.cpp:144][NNOP][Tiling][2165043] errno[561000] OpName:[aclnnFlashAttentionScore_309_FlashAttentionScore] Tiling failed [ERROR] OP(2162556,python3.1):2025-08-07-19:39:55.937.760 [kernel_workspace.cpp:730][NNOP][GetWorkspace][2165043] errno[561103] OpName:[aclnnFlashAttentionScore_309_FlashAttentionScore] Tiling Failed. [ERROR] OP(2162556,python3.1):2025-08-07-19:39:55.937.806 [kernel_workspace.cpp:102][NNOP][GetWorkspace][2165043] errno[561103] OpName:[aclnnFlashAttentionScore_309_FlashAttentionScore] Kernel GetWorkspace failed. opType: 42 [ERROR] OP(2162556,python3.1):2025-08-07-19:39:56.038.730 [flash_attention_score_tiling_general.cpp:1496][OP_TILING][AnalyzeOptionalInput][2165043] OpName:[FlashAttentionScore:FlashAttentionScore] "atten mask dim should be 2 or 4, but got 3" [ERROR] OP(2162556,python3.1):2025-08-07-19:39:56.038.917 [flash_attention_score_tiling_general.cpp:847][OP_TILING][GetShapeAttrsInfo][2165043] OpName:[FlashAttentionScore] "fail to analyze context info." [ERROR] OP(2162556,python3.1):2025-08-07-19:39:56.038.969 [kernel_workspace.cpp:144][NNOP][Tiling][2165043] errno[561000] OpName:[aclnnFlashAttentionScore_309_FlashAttentionScore] Tiling failed [ERROR] OP(2162556,python3.1):2025-08-07-19:39:56.039.037 [op_executor.cpp:617][NNOP][Launch][2165043] errno[561103] OpName:[aclnnFlashAttentionScore_309_FlashAttentionScore] Tiling Failed. [ERROR] OP(2162556,python3.1):2025-08-07-19:39:56.039.076 [op_executor.cpp:69][NNOP][Run][2165043] errno[561103] OpName:[aclnnFlashAttentionScore_309_FlashAttentionScore] Kernel Run failed. opType: 42, FlashAttentionScore [ERROR] OP(2162556,python3.1):2025-08-07-19:39:56.039.130 [op_executor.cpp:828][NNOP][Run][2165043] errno[561103] OpName:[aclnnFlashAttentionScore_309_FlashAttentionScore] launch failed for FlashAttentionScore, errno:561103. INFO: 127.0.0.1:59860 - "POST /v1/chat/completions HTTP/1.1" 500 Internal Server Error ERROR: Exception in ASGI application Traceback (most recent call last): File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/uvicorn/protocols/http/h11_impl.py", line 403, in run_asgi result = await app( # type: ignore[func-returns-value] File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/uvicorn/middleware/proxy_headers.py", line 60, in __call__ return await self.app(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/fastapi/applications.py", line 1054, in __call__ await super().__call__(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/applications.py", line 113, in __call__ await self.middleware_stack(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/errors.py", line 186, in __call__ raise exc File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/errors.py", line 164, in __call__ await self.app(scope, receive, _send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/cors.py", line 85, in __call__ await self.app(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/middleware/exceptions.py", line 63, in __call__ await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app raise exc File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app await app(scope, receive, sender) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 716, in __call__ await self.middleware_stack(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 736, in app await route.handle(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 290, in handle await self.app(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 78, in app await wrap_app_handling_exceptions(app, request)(scope, receive, send) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app raise exc File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/_exception_handler.py", line 42, in wrapped_app await app(scope, receive, sender) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/starlette/routing.py", line 75, in app response = await f(request) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/fastapi/routing.py", line 302, in app raw_response = await run_endpoint_function( File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/fastapi/routing.py", line 213, in run_endpoint_function return await dependant.call(**values) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/api/app.py", line 110, in create_chat_completion return await create_chat_completion_response(request, chat_model) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/api/chat.py", line 189, in create_chat_completion_response responses = await chat_model.achat( File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/chat/chat_model.py", line 92, in achat return await self.engine.chat(messages, system, tools, images, videos, audios, **input_kwargs) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/chat/hf_engine.py", line 363, in chat return await asyncio.to_thread(self._chat, *input_args) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/asyncio/threads.py", line 25, in to_thread return await loop.run_in_executor(None, func_call) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/concurrent/futures/thread.py", line 52, in run result = self.fn(*self.args, **self.kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) File "/home/ma-user/work/LLaMA-Factory/src/llamafactory/chat/hf_engine.py", line 240, in _chat generate_output = model.generate(**gen_kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context return func(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/generation/utils.py", line 2460, in generate result = self._sample( File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/generation/utils.py", line 3426, in _sample outputs = self(**model_inputs, return_dict=True) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py", line 1757, in forward image_embeds = self.visual(pixel_values, grid_thw=image_grid_thw) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py", line 546, in forward hidden_states = blk(hidden_states, cu_seqlens=cu_seqlens_now, position_embeddings=position_embeddings) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py", line 339, in forward hidden_states = hidden_states + self.attn( File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl return forward_call(*args, **kwargs) File "/home/ma-user/anaconda3/envs/llamafactory/lib/python3.10/site-packages/transformers/models/qwen2_5_vl/modeling_qwen2_5_vl.py", line 306, in forward attn_output = F.scaled_dot_product_attention( RuntimeError: call aclnnFlashAttentionScore failed, detail:EZ9999: Inner Error! EZ9999: [PID: 2162556] 2025-08-07-19:39:55.937.270 atten mask dim should be 2 or 4, but got 3[FUNC:AnalyzeOptionalInput][FILE:flash_attention_score_tiling_general.cpp][LINE:1496] TraceBack (most recent call last): fail to analyze context info.[FUNC:GetShapeAttrsInfo][FILE:flash_attention_score_tiling_general.cpp][LINE:847] Tiling failed Tiling Failed. Kernel GetWorkspace failed. opType: 42 Kernel Run failed. opType: 42, FlashAttentionScore launch failed for FlashAttentionScore, errno:561103. [ERROR] 2025-08-07-19:39:56 (PID:2162556, Device:0, RankID:-1) ERR01100 OPS call acl api failed

2025年07月版本 <ModelArts Versatile-AI原生应用引擎 体验入口>华为开发者空间--开发平台--AI Agent （请在PC端点击进入） Agent编排中心 · 我的工作流01 工作流支持多模态大模型的接入节点在创建工作流时，大模型节点从LLM（大语言模型）拓展至VLM（视觉语音大模型）/LMM（多模态大模型） 业务价值：让多模态模型在视频/图片理解、摘要分析等场景上进行处理，整合多种信息源，以更全面、准确的方式理解信息，丰富工作流的构建方式，深化工作流的设计科学度与执行精准度。  · 我的Agent 02 支持智能体中定期触发任务的能力在Agent开发过程中添加触发器，按照触发器的设置使Agent定时执行任务，自主规划/工作流模式下，定时任务按分钟/小时间隔触发；复杂任务规划模式下，可自主配置触发时间设置任务。 业务价值：实现智能化工作流程，更好地适配业务场景，训练AI助手自主开展任务，保障实时信息刷新，支撑连续性且高效的服务。  03 复杂任务规划模式下支持知识库在创建单Agent（复杂任务规划）模式下，可为Agent添加知识库 业务价值：丰富Agent任务规划数据来源，可以使Agent进行更专业的信息输入，扩大相关数据交互范围，增强回答的专业可信度，提升任务生成质量，为复杂任务的精准稳定运行护航。  模型中心 · 我的模型服务 04 支持接入华为云MaaS大模型支持从Modelarts大模型开发平台自动拉取大模型，如当前免费的DeepSeek-V3、R1，接入作为智能体的模型服务。 业务价值：打通开发者用户在MaaS服务下的模型，实现灵活调用，简化操作，提高模型利用率。    工程中心 · 项目管理 05 新增项目功能，支持项目导入导出项目是用来将一组相关的元素进行统一管理，其中元素包括Agent、工具、MCP、工作流、知识库、提示语、模型服务等。工程中心-项目管理，用户可对当前租户下的项目导出为本地压缩文件，同时支持一键导入已导出的项目。 业务价值：面对多租户切换的需求，帮助用户将前期已经开发完成的Agent、工具、工作流、知识库等实现快速迁移，满足多租户无痛迁移，适用多业务场景，通过导出，方便将该项目迁移到不同的环境中；通过导入，帮助用户快速创建复杂的智能应用。  APIModelArts Versatile-AI原生应用引擎调用接口支持AK/SK、API Key或Token认证鉴权，进行身份认证与快速调用。 · 应用中心 06 提供对话式智能体的增删改查接口涵盖创建Agent、更新Agent、根据ID查询Agent、删除指定Agent4项，支持API模式下的快速操作。创建：创建一个新的智能体。调用此接口创建一个智能体后，此智能体为未发布的草稿状态，创建后可以在Versatile智能体列表中查看智能体。更新：通过此API可更新通过Versatile平台或API方式创建的所有智能体。查询：根据智能体的ID查询指定的Agent，包含草稿状态的智能体和已发布的智能体。删除：根据Agent智能体的ID删除指定的Agent。 07 提供查询MCP列表的接口提供API接口给开发者空间查询MCP资产，接口符合开发者合约规范。通过传入的limit参数查询资产中心MCP列表。  · 知识中心 08 平台开放接口支持IAM鉴权方式开放Agent接口，支持公共知识库和个人知识库，支持IAM鉴权。涵盖查询知识库详情/更新知识库/查询知识库列表（Token认证）、查询知识数据集列表（Token认证）接口。   点击可前往>>华为云ModelArts Versatile-AI原生应用引擎官网

体验华为开发者空间《基于华为开发者空间云开发环境部署Coze Studio + Maas构建智能体应用》案例，反馈改进建议，请直接在评论区反馈即可。体验指导：https://devstation.connect.huaweicloud.com/space/devportal/casecenter/ce2cabd01f704b3c8bbc91b91c312545/1