ATC转换AMCT量化后的模型失败,EZ3002: Optype [AscendQuant] of Ops kernel [AIcoreEngine] is unsupported.
收藏回复举报
ATC转换AMCT量化后的模型失败,EZ3002: Optype [AscendQuant] of Ops kernel [AIcoreEngine] is unsupported.
t('forum.solved') 已解决
新人帖
发表于2024-12-02 16:56:04
0 查看

【问题详细描述】

  • 报错截图

(报错图片上传一直失败,用bash代码块代替)

(base) [root@localhost amct_pytorch]# atc --model=/tmp/pycharm_project_842/optimizer/quantize/result/model_deploy_model.onnx --framework=5 --output=/tmp/pycharm_project_842/optimizer/quantize/result/ATC --soc_version=Ascend910B3 --precision_mode_v2=origin --log=debug
ATC start working now, please wait for a moment.
...
ATC run failed, Please check the detail log, Try 'atc --help' for more information
EZ3002: [PID: 3542304] 2024-12-02-16:33:13.753.940 Optype [AscendQuant] of Ops kernel [AIcoreEngine] is unsupported. Reason: [tbe-custom]:op type AscendQuant is not found in this op store.[tbe-custom]:op type AscendQuant is not found in this op store.The op is dynamic shape, but is not configured to support dynamic shape in op store.
        Possible Cause: The operator type is unsupported in the operator information library due to specification mismatch.
        Solution: Submit an issue to request for support at https://gitee.com/ascend, or remove this type of operators from your model.
        TraceBack (most recent call last):
        No supported Ops kernel and engine are found for [transformer_ue.0.attention.ln.quant], optype [AscendQuant].
        Assert ((SelectEngine(node_ptr, exclude_engines, is_check_support_success, op_info)) == ge::SUCCESS) failed[FUNC:operator()][FILE:engine_place.cc][LINE:149]
        No supported Ops kernel and engine are found for [transformer_ue.0.feed_forward.linear_1.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_ue.0.feed_forward.linear_2.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.0.attention.ln.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.0.feed_forward.linear_1.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.0.feed_forward.linear_2.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_ue.1.attention.ln.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_ue.1.feed_forward.linear_1.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_ue.1.feed_forward.linear_2.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.1.attention.ln.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.1.feed_forward.linear_1.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.1.feed_forward.linear_2.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_ue.2.attention.ln.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_ue.2.feed_forward.linear_1.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_ue.2.feed_forward.linear_2.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.2.attention.ln.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.2.feed_forward.linear_1.quant], optype [AscendQuant].
        No supported Ops kernel and engine are found for [transformer_rbg.2.feed_forward.linear_2.quant], optype [AscendQuant].
        RunAllSubgraphs failed, graph=ATC.[FUNC:RunAllSubgraphs][FILE:engine_place.cc][LINE:123]
        build graph failed, graph id:0, ret:4294967295[FUNC:BuildModelWithGraphId][FILE:ge_generator.cc][LINE:1618]
  • 文字描述

对使用AMCT工具量化后的部署onnx模型进行格式转换,ATC转换报错如图所示:EZ3002: Optype [AscendQuant] of Ops kernel [AIcoreEngine]  is unsupported.

【问题相关文件】:根据官网文档上的进行安装跟量化:https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/800alpha002/devaids/devtools/amct/atlasamct_16_0047.html

【昇腾产品型号】:Atlas800 9000

【版本信息】:

  • CANN版本:8.0.0.alpha002

  • Python版本:Python 3.10.14

  • 系统版本: EulerOS release 2.0 (SP11)
    eulerversion=EulerOS_Server_V200R011C00SPC507B910_aarch64 kernelversion=5.10.0-60.18.0.50.h1083

【用户类型】:企业

【开发需求】

  • 功能需求: 使用AMCT工具量化transformer模型(pytorch框架)后加快推理时延

【应用目的】: 商业

本帖最后由 匿名用户2024/12/03 09:44:33 编辑

我要发帖子