incorporate #117 from redhat-et/ilab-on-ocp

Signed-off-by: Michael Clifford <[email protected]> Co-authored-by: Michael Clifford <[email protected]> Co-authored-by: Sébastien Han <[email protected]>
opendatahub-io · Oct 22, 2024 · cbf4a40 · cbf4a40
1 parent 1fb06b0
commit cbf4a40
Showing 1 changed file with 21 additions and 5 deletions.
diff --git a/instructlab/standalone/standalone.py b/instructlab/standalone/standalone.py
@@ -1550,6 +1550,7 @@ def data_processing(train_args: TrainingArgs) -> None:
 def create_eval_job(
     namespace: str,
     eval_type: str,
+    judge_serving_model_secret: str,
     nproc_per_node: int = 1,
 ) -> kubernetes.client.V1Job:
     """
@@ -1560,6 +1561,7 @@ def create_eval_job(
     Args:
         namespace (str): The namespace in which the job will be created.
         eval_type (str): The type of evaluation to run.
+        judge_serving_model_secret (str): The name of the Kubernetes Secret containing the judge
         nproc_per_node (int): The number of processes per node.
 
     Returns:
@@ -1729,7 +1731,7 @@ def shutdown_vllm(process: subprocess.Popen, timeout: int = 20):
         max_workers = usable_cpu_count
 
     # modify model_list to ignore any jsonl files present in the directory
-    models_list = [model for model in models_list if model.endswith(".jsonl") != True]
+    models_list = [model for model in models_list if not model.endswith(".jsonl")]
     for model_name in models_list:
         print(f"Serving candidate model: {model_name}")
         model_path = f"{models_path_prefix}/{model_name}"
@@ -2275,7 +2277,7 @@ def find_node_dataset_directories(base_dir: str):
                 env_from=[
                     kubernetes.client.V1EnvFromSource(
                         secret_ref=kubernetes.client.V1SecretEnvSource(
-                            name=JUDGE_SERVING_NAME
+                            name=judge_serving_model_secret
                         )
                     ),
                 ],
@@ -2310,7 +2312,7 @@ def find_node_dataset_directories(base_dir: str):
                 env_from=[
                     kubernetes.client.V1EnvFromSource(
                         secret_ref=kubernetes.client.V1SecretEnvSource(
-                            name=JUDGE_SERVING_NAME
+                            name=judge_serving_model_secret
                         )
                     ),
                 ],
@@ -2854,6 +2856,9 @@ def decode_base64(data):
                         f"Secret {judge_serving_model_secret} not found in namespace {namespace}."
                     ) from exc
 
+    # Set the judge secret in the context for the evaluation job
+    ctx.obj["judge_serving_model_secret"] = judge_serving_model_secret
+
     # list of PVCs to create and their details
     pvcs = [
         {
@@ -3112,6 +3117,13 @@ def evaluation(ctx: click.Context) -> str:
     namespace = ctx.obj["namespace"]
     eval_type = ctx.obj["eval_type"]
     dry_run = ctx.obj["dry_run"]
+    judge_serving_model_secret = ctx.obj["judge_serving_model_secret"]
+
+    # This should only happen if the script is called with the "evaluation" subcommand
+    if not judge_serving_model_secret:
+        raise ValueError(
+            "Judge serving model secret must be provided with --judge-serving-model-secret."
+        )
 
     if eval_type is None:
         raise ValueError(
@@ -3121,7 +3133,11 @@ def evaluation(ctx: click.Context) -> str:
     logger.info("Running %s evaluation.", eval_type)
 
     # Create and run the evaluation job
-    job = create_eval_job(namespace=namespace, eval_type=eval_type)
+    job = create_eval_job(
+        namespace=namespace,
+        eval_type=eval_type,
+        judge_serving_model_secret=judge_serving_model_secret,
+    )
 
     if dry_run:
         logger.info("Dry run: Job would be created.\n%s", job)
@@ -3196,4 +3212,4 @@ def upload_trained_model(ctx: click.Context):
         logger.info("Failed to load kube config. Trying in-cluster config")
         kubernetes.config.load_incluster_config()
 
-    cli()
+    cli()