Save-best, Save-best-rule 컬럼 varchar100으로 변경

Reviewed-on: #185

src/main/java/com/kamco/cd/training/postgres/entity/ModelHyperParamEntity.java

@@ -181,15 +181,15 @@ public class ModelHyperParamEntity {
   private String metrics = "mFscore,mIoU";
 
   /** Default: changed_fscore */
-  @Size(max = 30)
+  @Size(max = 100)
   @NotNull
-  @Column(name = "save_best", nullable = false, length = 30)
+  @Column(name = "save_best", nullable = false, length = 100)
   private String saveBest = "changed_fscore";
 
   /** Default: greater */
-  @Size(max = 10)
+  @Size(max = 100)
   @NotNull
-  @Column(name = "save_best_rule", nullable = false, length = 10)
+  @Column(name = "save_best_rule", nullable = false, length = 100)
   private String saveBestRule = "greater";
 
   /** Default: 1 */

src/main/java/com/kamco/cd/training/train/service/DockerTrainService.java

@@ -56,6 +56,13 @@ public class DockerTrainService {
   @Value("${spring.profiles.active}")
   private String profile;
 
+  @Value("${hyper.parameter.gpus}")
+  private String hyperGpus;
+
+
+  @Value("${hyper.parameter.gpu-ids}")
+  private String hyperGpuIds;
+
   private final ModelTrainJobCoreService modelTrainJobCoreService;
 
   /**
@@ -285,11 +292,13 @@ public class DockerTrainService {
     //    addArg(c, "--gpu-ids", req.getGpuIds()); // null
     if ("prod".equals(profile)) {
       addArg(c, "--batch-size", 2); // 학습서버 GPU 1개인 곳은 batch-size:2 까지만 가능
-      addArg(c, "--gpus", "1"); // 학습서버 GPU 1개인 곳은 1이어야 함
-      addArg(c, "--gpu-ids", "0"); // 학습서버 GPU 1개인 곳은 0이어야 함
+
     } else {
       addArg(c, "--batch-size", req.getBatchSize()); // 학습서버 GPU 1개인 곳은 batch-size:2 까지만 가능
     }
+    addArg(c, "--gpus", hyperGpus); // 학습서버 GPU 1개인 곳은 1이어야 함
+    addArg(c, "--gpu-ids", hyperGpuIds); // 학습서버 GPU 1개인 곳은 0이어야 함
+
     addArg(c, "--lr", req.getLearningRate());
     addArg(c, "--backbone", req.getBackbone());
     addArg(c, "--epochs", req.getEpochs());

src/main/resources/application-prod.yml

@@ -41,3 +41,7 @@ train:
     container_prefix: kamco-cd-train
     shm_size: 16g
     ipc_host: true
+hyper:
+  parameter:
+    gpus: 4
+    gpu-ids: 0,1,2,3

src/main/resources/application.yml

@@ -78,3 +78,8 @@ management:
       exposure:
         include:
           - "health"
+hyper:
+  parameter:
+    gpus: 1
+    gpu-ids: 0
+