Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-09-20 18:45:09 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Include trust remote code usage in openai api's embedder (#4513)

Browse source
This commit is contained in:
MrMojoR 2023-11-08 15:25:43 +01:00 committed by GitHub
parent 6c7aad11f3
commit 1754a3761b
WARNING! Although there is a key with this ID in the database it does not verify this commit! This commit is SUSPICIOUS.
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 11 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

18
extensions/openai/embeddings.py
View file

@ -3,7 +3,9 @@ import os
import numpy as np
from extensions.openai.errors import ServiceUnavailableError
from extensions.openai.utils import debug_msg, float_list_to_base64
from sentence_transformers import SentenceTransformer
from transformers import AutoModel
from modules import shared
embeddings_params_initialized = False
@ -26,21 +28,23 @@ def initialize_embedding_params():
embeddings_params_initialized = True
def load_embedding_model(model: str) -> SentenceTransformer:
def load_embedding_model(model: str):
initialize_embedding_params()
global embeddings_device, embeddings_model
try:
print(f"Try embedding model: {model} on {embeddings_device}")
# see: https://www.sbert.net/docs/package_reference/SentenceTransformer.html#sentence_transformers.SentenceTransformer
embeddings_model = SentenceTransformer(model, device=embeddings_device)
# ... embeddings_model.device doesn't seem to work, always cpu anyways? but specify cpu anyways to free more VRAM
print(f"\nLoaded embedding model: {model} on {embeddings_model.device} [always seems to say 'cpu', even if 'cuda'], max sequence length: {embeddings_model.max_seq_length}")
trust = shared.args.trust_remote_code
if embeddings_device == 'cpu':
embeddings_model = AutoModel.from_pretrained(model, trust_remote_code=trust).to("cpu", dtype=float)
else: #use the auto mode
embeddings_model = AutoModel.from_pretrained(model, trust_remote_code=trust)
print(f"\nLoaded embedding model: {model} on {embeddings_model.device}")
except Exception as e:
embeddings_model = None
raise ServiceUnavailableError(f"Error: Failed to load embedding model: {model}", internal_message=repr(e))
def get_embeddings_model() -> SentenceTransformer:
def get_embeddings_model() -> AutoModel:
initialize_embedding_params()
global embeddings_model, st_model
if st_model and not embeddings_model: