highperfocused/danswer
1
0
Fork 0
mirror of https://github.com/danswer-ai/danswer.git synced 2025-09-26 11:58:28 +02:00
Code Issues Packages Projects Releases Wiki Activity

Added env var to skip warm up (#3633)

Browse Source
This commit is contained in:
hagen-danswer
2025-01-08 14:29:15 -08:00
committed by GitHub
parent d7bc32c0ec
commit 0f9842064f
3 changed files with 7 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.vscode/env_template.txt vendored
View File

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

1
backend/onyx/configs/app_configs.py
View File

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

4
backend/onyx/natural_language_processing/search_nlp_models.py
View File

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