highperfocused/danswer
1
0
Fork 0
mirror of https://github.com/danswer-ai/danswer.git synced 2025-07-12 22:23:01 +02:00
Code Issues Packages Projects Releases Wiki Activity

Fix Model Server (#1320)

Browse Source
This commit is contained in:
Yuhong Sun
2024-04-10 23:13:22 -07:00
committed by GitHub
parent f346c2fc86
commit b59912884b
18 changed files with 48 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
backend/model_server/custom_models.py
View File

@ -8,11 +8,11 @@ from transformers import TFDistilBertForSequenceClassification
from model_server.constants import MODEL_WARM_UP_STRING
from model_server.utils import simple_log_function_time
from shared_configs.configs import INDEXING_ONLY
from shared_configs.configs import INTENT_MODEL_CONTEXT_SIZE
from shared_configs.configs import INTENT_MODEL_VERSION
from shared_configs.model_server_models import IntentRequest
from shared_configs.model_server_models import IntentResponse
from shared_configs.nlp_model_configs import INDEXING_ONLY
from shared_configs.nlp_model_configs import INTENT_MODEL_CONTEXT_SIZE
from shared_configs.nlp_model_configs import INTENT_MODEL_VERSION
router = APIRouter(prefix="/custom")

6
backend/model_server/encoders.py
View File

@ -9,13 +9,13 @@ from sentence_transformers import SentenceTransformer # type: ignore
from danswer.utils.logger import setup_logger
from model_server.constants import MODEL_WARM_UP_STRING
from model_server.utils import simple_log_function_time
from shared_configs.configs import CROSS_EMBED_CONTEXT_SIZE
from shared_configs.configs import CROSS_ENCODER_MODEL_ENSEMBLE
from shared_configs.configs import INDEXING_ONLY
from shared_configs.model_server_models import EmbedRequest
from shared_configs.model_server_models import EmbedResponse
from shared_configs.model_server_models import RerankRequest
from shared_configs.model_server_models import RerankResponse
from shared_configs.nlp_model_configs import CROSS_EMBED_CONTEXT_SIZE
from shared_configs.nlp_model_configs import CROSS_ENCODER_MODEL_ENSEMBLE
from shared_configs.nlp_model_configs import INDEXING_ONLY
logger = setup_logger()

12
backend/model_server/main.py
View File

@ -8,17 +8,17 @@ from fastapi import FastAPI
from transformers import logging as transformer_logging # type:ignore
from danswer import __version__
from danswer.configs.app_configs import MODEL_SERVER_ALLOWED_HOST
from danswer.configs.app_configs import MODEL_SERVER_PORT
from danswer.utils.logger import setup_logger
from model_server.custom_models import router as custom_models_router
from model_server.custom_models import warm_up_intent_model
from model_server.encoders import router as encoders_router
from model_server.encoders import warm_up_cross_encoders
from shared_configs.nlp_model_configs import ENABLE_RERANKING_ASYNC_FLOW
from shared_configs.nlp_model_configs import ENABLE_RERANKING_REAL_TIME_FLOW
from shared_configs.nlp_model_configs import INDEXING_ONLY
from shared_configs.nlp_model_configs import MIN_THREADS_ML_MODELS
from shared_configs.configs import ENABLE_RERANKING_ASYNC_FLOW
from shared_configs.configs import ENABLE_RERANKING_REAL_TIME_FLOW
from shared_configs.configs import INDEXING_ONLY
from shared_configs.configs import MIN_THREADS_ML_MODELS
from shared_configs.configs import MODEL_SERVER_ALLOWED_HOST
from shared_configs.configs import MODEL_SERVER_PORT
os.environ["TOKENIZERS_PARALLELISM"] = "false"
os.environ["HF_HUB_DISABLE_TELEMETRY"] = "1"