holetron-lab/fleet-memory
1
0
Fork 0
Code Releases Packages Activity Actions

fix: hide hf logging (#295)

Browse source
This commit is contained in:
Nicolò Boschi 2026-02-04 13:12:30 +01:00 committed by GitHub
parent 9a776e9f58
commit 34d9188e13
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 54 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
docker/standalone/Dockerfile
View file

@ -208,6 +208,10 @@ ENV HINDSIGHT_API_LOG_LEVEL=info
ENV HINDSIGHT_ENABLE_API=true
ENV HINDSIGHT_ENABLE_CP=false
ENV PYTHONUNBUFFERED=1
# Suppress verbose transformers/HuggingFace model loading warnings
ENV TRANSFORMERS_VERBOSITY=error
ENV HF_HUB_VERBOSITY=error
ENV TOKENIZERS_PARALLELISM=false
CMD ["/app/start-all.sh"]
@ -336,6 +340,10 @@ ENV HINDSIGHT_CP_DATAPLANE_API_URL=http://localhost:8888
ENV HINDSIGHT_ENABLE_API=true
ENV HINDSIGHT_ENABLE_CP=true
ENV PYTHONUNBUFFERED=1
# Suppress verbose transformers/HuggingFace model loading warnings
ENV TRANSFORMERS_VERBOSITY=error
ENV HF_HUB_VERBOSITY=error
ENV TOKENIZERS_PARALLELISM=false
CMD ["/app/start-all.sh"]

28
hindsight-api/hindsight_api/engine/cross_encoder.py
View file

@ -9,6 +9,7 @@ Configuration via environment variables - see hindsight_api.config for all env v
import asyncio
import logging
import os
import warnings
from abc import ABC, abstractmethod
from concurrent.futures import ThreadPoolExecutor
@ -162,11 +163,28 @@ class LocalSTCrossEncoder(CrossEncoderModel):
except Exception as e:
logger.warning(f"Failed to detect GPU/MPS, falling back to CPU: {e}")
self._model = CrossEncoder(
self.model_name,
device=device,
model_kwargs={"low_cpu_mem_usage": False},
)
# Suppress verbose transformers warnings during model loading
# This suppresses the "UNEXPECTED" warnings from CrossEncoder which are harmless
# but look alarming to users (e.g., "embeddings.position_ids | UNEXPECTED")
with warnings.catch_warnings():
warnings.filterwarnings("ignore", category=UserWarning)
warnings.filterwarnings("ignore", message=".*was not found in model state dict.*")
warnings.filterwarnings("ignore", message=".*UNEXPECTED.*")
# Also suppress transformers library logging temporarily
transformers_logger = logging.getLogger("transformers")
original_level = transformers_logger.level
transformers_logger.setLevel(logging.ERROR)
try:
self._model = CrossEncoder(
self.model_name,
device=device,
model_kwargs={"low_cpu_mem_usage": False},
)
finally:
# Restore original logging level
transformers_logger.setLevel(original_level)
# Initialize shared executor (limited workers naturally limits concurrency)
if LocalSTCrossEncoder._executor is None:

28
hindsight-api/hindsight_api/engine/embeddings.py
View file

@ -11,6 +11,7 @@ Configuration via environment variables - see hindsight_api.config for all env v
import logging
import os
import warnings
from abc import ABC, abstractmethod
import httpx
@ -157,11 +158,28 @@ class LocalSTEmbeddings(Embeddings):
except Exception as e:
logger.warning(f"Failed to detect GPU/MPS, falling back to CPU: {e}")
self._model = SentenceTransformer(
self.model_name,
device=device,
model_kwargs={"low_cpu_mem_usage": False},
)
# Suppress verbose transformers warnings during model loading
# This suppresses the "UNEXPECTED" warnings from BertModel which are harmless
# but look alarming to users (e.g., "embeddings.position_ids | UNEXPECTED")
with warnings.catch_warnings():
warnings.filterwarnings("ignore", category=UserWarning)
warnings.filterwarnings("ignore", message=".*was not found in model state dict.*")
warnings.filterwarnings("ignore", message=".*UNEXPECTED.*")
# Also suppress transformers library logging temporarily
transformers_logger = logging.getLogger("transformers")
original_level = transformers_logger.level
transformers_logger.setLevel(logging.ERROR)
try:
self._model = SentenceTransformer(
self.model_name,
device=device,
model_kwargs={"low_cpu_mem_usage": False},
)
finally:
# Restore original logging level
transformers_logger.setLevel(original_level)
self._dimension = self._model.get_sentence_embedding_dimension()
logger.info(f"Embeddings: local provider initialized (dim: {self._dimension})")