You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

dubhe-train.yaml 1.8 kB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647
  1. #训练常量
  2. train-job:
  3. namespace: "namespace-"
  4. separator: "-"
  5. version-label: "v"
  6. pod-name: "platform/resource-name"
  7. python-format: " --"
  8. manage: "train-manage"
  9. model-path: "model-out"
  10. train-model-path-value: "${train-job.docker-train-path}/${train-job.model-path}"
  11. train-out-path-value: "${train-job.docker-train-path}/${train-job.out-path}"
  12. visualized-log-path-value: "${train-job.docker-train-path}/${train-job.visualized-log-path}"
  13. out-path: "out"
  14. load-path: "load"
  15. load-key: "model_load_dir"
  16. load-val-dataset-key: "val_data_url"
  17. visualized-log-path: "visualizedlog"
  18. docker-dataset-path: "/dataset"
  19. docker-train-path: "/workspace"
  20. docker-model-path: "/modeldir"
  21. docker-val-dataset-path: "/valdataset"
  22. docker-train-model-path: "train_model_out=${train-job.docker-train-path}/${train-job.model-path}"
  23. docker-train-out-path: "train_out=${train-job.docker-train-path}/${train-job.out-path}"
  24. docker-visualized-log-path: "train_visualized_log=${train-job.docker-train-path}/${train-job.visualized-log-path}"
  25. docker-dataset: "data_url=${train-job.docker-dataset-path}"
  26. docker-teacher-model-key: "teacher_path_list"
  27. docker-teacher-model-path: "/model/teacher"
  28. docker-student-model-key: "student_path_list"
  29. docker-student-model-path: "/model/student"
  30. docker-atlas-dataset-key: "atlas_dataset_paths"
  31. docker-atlas-val-dataset-key: "atlas_val_dataset_paths"
  32. atlas-anaconda: "cd /root/anaconda3/bin && source activate base"
  33. atlas-pythonioencoding: "PYTHONIOENCODING=utf-8 "
  34. eight: "8"
  35. plus-eight: "+8"
  36. node-ips: "node_ips"
  37. node-num: "num_nodes"
  38. gpu-num-per-node: "gpu_num_per_node"
  39. #关闭熔断机制
  40. feign:
  41. hystrix:
  42. enabled: false
  43. # 可匿名访问路径
  44. security:
  45. permitAll:
  46. matchers: /api/k8s/callback/pod/algorithm

一站式算法开发平台、高性能分布式深度学习框架、先进算法模型库、视觉模型炼知平台、数据可视化分析平台等一系列平台及工具,在模型高效分布式训练、数据处理和可视分析、模型炼知和轻量化等技术上形成独特优势,目前已在产学研等各领域近千家单位及个人提供AI应用赋能