Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix fastembed reloading issue. #4117

Merged
merged 3 commits into from
Dec 19, 2024
Merged

Fix fastembed reloading issue. #4117

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
499f95f
Fix fastembed reloading issue.
KevinHuSh Dec 19, 2024
bc3fe78
Trival.
KevinHuSh Dec 19, 2024
9fefb8e
Trival.
KevinHuSh Dec 19, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 19 additions & 2 deletions rag/llm/embedding_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,6 +47,7 @@ def encode_queries(self, text: str):

class DefaultEmbedding(Base):
_model = None
_model_name = ""
_model_lock = threading.Lock()
def __init__(self, key, model_name, **kwargs):
"""
Expand All @@ -69,6 +70,7 @@ def __init__(self, key, model_name, **kwargs):
DefaultEmbedding._model = FlagModel(os.path.join(get_home_cache_dir(), re.sub(r"^[a-zA-Z0-9]+/", "", model_name)),
query_instruction_for_retrieval="为这个句子生成表示以用于检索相关文章：",
use_fp16=torch.cuda.is_available())
DefaultEmbedding._model_name = model_name
except Exception:
model_dir = snapshot_download(repo_id="BAAI/bge-large-zh-v1.5",
local_dir=os.path.join(get_home_cache_dir(), re.sub(r"^[a-zA-Z0-9]+/", "", model_name)),
Expand All @@ -77,6 +79,7 @@ def __init__(self, key, model_name, **kwargs):
query_instruction_for_retrieval="为这个句子生成表示以用于检索相关文章：",
use_fp16=torch.cuda.is_available())
self._model = DefaultEmbedding._model
self._model_name = DefaultEmbedding._model_name

def encode(self, texts: list):
batch_size = 16
Expand Down Expand Up @@ -250,6 +253,8 @@ def encode_queries(self, text):

class FastEmbed(Base):
_model = None
_model_name = ""
_model_lock = threading.Lock()

def __init__(
self,
Expand All @@ -260,8 +265,20 @@ def __init__(
**kwargs,
):
if not settings.LIGHTEN and not FastEmbed._model:
from fastembed import TextEmbedding
self._model = TextEmbedding(model_name, cache_dir, threads, **kwargs)
with FastEmbed._model_lock:
from fastembed import TextEmbedding
if not FastEmbed._model or model_name != FastEmbed._model_name:
try:
FastEmbed._model = TextEmbedding(model_name, cache_dir, threads, **kwargs)
FastEmbed._model_name = model_name
except Exception:
cache_dir = snapshot_download(repo_id="BAAI/bge-small-en-v1.5",
local_dir=os.path.join(get_home_cache_dir(),
re.sub(r"^[a-zA-Z0-9]+/", "", model_name)),
local_dir_use_symlinks=False)
FastEmbed._model = TextEmbedding(model_name, cache_dir, threads, **kwargs)
self._model = FastEmbed._model
self._model_name = model_name

def encode(self, texts: list):
# Using the internal tokenizer to encode the texts and get the total
Expand Down