From 1c7877b048d2131c8c133c26c60c8ccd342f0b0c Mon Sep 17 00:00:00 2001 From: Shota Totsuka <153569547+totsukash@users.noreply.github.com> Date: Tue, 24 Sep 2024 21:53:26 +0900 Subject: [PATCH] fix: remove harm category setting from vertex ai (#8721) --- .../model_providers/vertex_ai/llm/llm.py | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-) diff --git a/api/core/model_runtime/model_providers/vertex_ai/llm/llm.py b/api/core/model_runtime/model_providers/vertex_ai/llm/llm.py index da69b7cdf3..1dd785d545 100644 --- a/api/core/model_runtime/model_providers/vertex_ai/llm/llm.py +++ b/api/core/model_runtime/model_providers/vertex_ai/llm/llm.py @@ -2,6 +2,7 @@ import base64 import io import json import logging +import time from collections.abc import Generator from typing import Optional, Union, cast @@ -20,7 +21,6 @@ from google.api_core import exceptions from google.cloud import aiplatform from google.oauth2 import service_account from PIL import Image -from vertexai.generative_models import HarmBlockThreshold, HarmCategory from core.model_runtime.entities.llm_entities import LLMResult, LLMResultChunk, LLMResultChunkDelta, LLMUsage from core.model_runtime.entities.message_entities import ( @@ -34,6 +34,7 @@ from core.model_runtime.entities.message_entities import ( ToolPromptMessage, UserPromptMessage, ) +from core.model_runtime.entities.model_entities import PriceType from core.model_runtime.errors.invoke import ( InvokeAuthorizationError, InvokeBadRequestError, @@ -503,20 +504,12 @@ class VertexAiLargeLanguageModel(LargeLanguageModel): else: history.append(content) - safety_settings = { - HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE, - HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE, - HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE, - HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE, - } - google_model = glm.GenerativeModel(model_name=model, system_instruction=system_instruction) response = google_model.generate_content( contents=history, generation_config=glm.GenerationConfig(**config_kwargs), stream=stream, - safety_settings=safety_settings, tools=self._convert_tools_to_glm_tool(tools) if tools else None, )