Added env var to skip warm up (#3633)

.vscode/env_template.txt

@@ -5,6 +5,8 @@
 # For local dev, often user Authentication is not needed
 AUTH_TYPE=disabled
 
+# Skip warm up for dev
+SKIP_WARM_UP=True
 
 # Always keep these on for Dev
 # Logs all model prompts to stdout

backend/onyx/configs/app_configs.py

@@ -17,6 +17,7 @@ APP_PORT = 8080
 # prefix from requests directed towards the API server. In these cases, set this to `/api`
 APP_API_PREFIX = os.environ.get("API_PREFIX", "")
 
+SKIP_WARM_UP = os.environ.get("SKIP_WARM_UP", "").lower() == "true"
 
 #####
 # User Facing Features Configs

backend/onyx/natural_language_processing/search_nlp_models.py

@@ -12,6 +12,7 @@ from requests import Response
 from retry import retry
 
 from onyx.configs.app_configs import LARGE_CHUNK_RATIO
+from onyx.configs.app_configs import SKIP_WARM_UP
 from onyx.configs.model_configs import BATCH_SIZE_ENCODE_CHUNKS
 from onyx.configs.model_configs import (
     BATCH_SIZE_ENCODE_CHUNKS_FOR_API_EMBEDDING_SERVICES,
@@ -384,6 +385,9 @@ def warm_up_bi_encoder(
     embedding_model: EmbeddingModel,
     non_blocking: bool = False,
 ) -> None:
+    if SKIP_WARM_UP:
+        return
+
     warm_up_str = " ".join(WARM_UP_STRINGS)
 
     logger.debug(f"Warming up encoder model: {embedding_model.model_name}")