机器学习PAI是只支持部署model.json文件吗？
机器学习PAI是只支持部署model.json文件吗？这边部署失败了 Traceback (most recent call last):
[2023-11-02 13:31:53] File "xgboost.py", line 74, in
[2023-11-02 13:31:53] runner.run()
[2023-11-02 13:31:53] File "/home/admin/docker_ml/workspace/predictor/loan_multiloan_correct_risk_model_v11/ENV/lib/python3.8/site-packages/allspark/init.py", line 896, in run
[2023-11-02 13:31:53] self.initialize()
[2023-11-02 13:31:53] File "xgboost.py", line 54, in initialize
[2023-11-02 13:31:53] tl2cgen.export_lib(model, toolchain='gcc', libpath='./mymodel.so',
[2023-11-02 13:31:53] File "/home/admin/docker_ml/workspace/predictor/loan_multiloan_correct_risk_model_v11/ENV/lib/python3.8/site-packages/tl2cgen/shortcuts.py", line 82, in export_lib
[2023-11-02 13:31:53] temp_libpath = create_shared(
[2023-11-02 13:31:53] File "/home/admin/docker_ml/workspace/predictor/loan_multiloan_correct_risk_model_v11/ENV/lib/python3.8/site-packages/tl2cgen/create_shared.py", line 105, in create_shared
[2023-11-02 13:31:53] libpath = _create_shared(
[2023-11-02 13:31:53] File "/home/admin/docker_ml/workspace/predictor/loan_multiloan_correct_risk_model_v11/ENV/lib/python3.8/site-packages/tl2cgen/contrib/gcc.py", line 68, in _create_shared_gcc
[2023-11-02 13:31:53] return _create_shared_base(dirpath, recipe, nthread=nthread, verbose=verbose)
[2023-11-02 13:31:53] File "/home/admin/docker_ml/workspace/predictor/loan_multiloan_correct_risk_model_v11/ENV/lib/python3.8/site-packages/tl2cgen/contrib/create_shared.py", line 98, in _create_shared_base
[2023-11-02 13:31:53] raise TL2cgenError(
[2023-11-02 13:31:53] tl2cgen.exception.TL2cgenError: Error occured in worker #0: gcc: internal compiler error: Killed (program cc1)
[2023-11-02 13:31:53] Please submit a full bug report,
[2023-11-02 13:31:53] with preprocessed source if appropriate.
[2023-11-02 13:31:53] See https://aone.alibaba-inc.com/project/767688/issue/ for instructions.
[2023-11-02 13:31:53] gcc: internal compiler error: Killed (program cc1)
[2023-11-02 13:31:53] Please submit a full bug report,
[2023-11-02 13:31:53] with preprocessed source if appropriate.
[2023-11-02 13:31:53] See https://aone.alibaba-inc.com/project/767688/issue/ for instructions.
[2023-11-02 13:31:53]
写回答
-
.json 和 .m 应该都可以，这个processor使用了treelite部署，不是原生的xgboost库，您如果有需求我可以临时打个包，具体需要使用哪个库，此回答整理自钉群“机器学习PAI交流群(答疑@值班)”
2023-11-08 07:37:46赞同 展开评论 打赏 -
面对过去，不要迷离；面对未来，不必彷徨；活在今天，你只要把自己完全展示给别人看从您的错误日志中可以看出，在运行Python脚本的过程中出现了GCC编译器的内部错误，这可能是由于程序资源不足或者编译时间过长造成的。建议您可以尝试以下解决方案：
- 检查您的程序，确保没有死循环或者其他消耗过多资源的情况；
- 在PAI上创建更大规格的集群，分配更多的计算资源；
- 重新安装GCC编译器，并尝试重新运行您的脚本。
如果问题仍然存在，请尝试联系阿里云技术支持团队寻求更多帮助和建议。
2023-11-06 13:26:46赞同 展开评论 打赏
版权声明：本文内容由阿里云实名注册用户自发贡献，版权归原作者所有，阿里云开发者社区不拥有其著作权，亦不承担相应法律责任。具体规则请查看《阿里云开发者社区用户服务协议》和《阿里云开发者社区知识产权保护指引》。如果您发现本社区中有涉嫌抄袭的内容，填写侵权投诉表单进行举报，一经查实，本社区将立刻删除涉嫌侵权内容。
