BadRequestError: Context Length Exceeded for Multiple Images in Base64 Conversion
When attempting to process multiple images with the following code, an error occurs indicating that the context length exceeds the model's maximum token limit. The specific error is:
BadRequestError: Error code: 400 - {'error': {'message': "This model's maximum context length is 128000 tokens. However, your messages resulted in 144641 tokens. Please reduce the length of the messages.", 'type': 'invalid_request_error', 'param': 'messages', 'code': 'context_length_exceeded'}}
Code:
when attempting to use multiple images, the following code results in an error:
`import base64
Function to convert image file to base64
def convert_image_to_base64(image_path):
with open(image_path, "rb") as image_file:
return base64.b64encode(image_file.read()).decode("utf-8")
Data with image paths
data = [
{
"title": "Step 1: Open the Application",
"content": [
{
"type": "image",
"path": "step1_image.png"
}
]
},
{
"title": "Step 2: Navigate to the Settings",
"content": [
{
"type": "image",
"path": "step2_image.png"
}
]
}
]
Function to construct message with base64 images
def construct_message(data):
message_parts = []
for item in data:
title = item["title"]
content_parts = []
for content in item["content"]:
if content["type"] == "image":
image_base64 = convert_image_to_base64(content["path"])
content_parts.append(f'<img src="data:image/png;base64,{image_base64}">')
message_parts.append(f"{title}\n" + "\n".join(content_parts))
return "\n\n".join(message_parts)
Construct the message with base64 image data
message = construct_message(data)
`
The agents are as follows:
`agent1 = MultimodalConversableAgent(
name="example-agent-1",
max_consecutive_auto_reply=10,
llm_config={"config_list": config_list_example, "temperature": 0.5, "max_tokens": 300},
system_message="You are an assistant who helps interpret and analyze image-based content.",
)
agent2 = AssistantAgent(
name="example-agent-2",
max_consecutive_auto_reply=10,
system_message="You are a support specialist who provides feedback on content interpretation.",
)
user_proxy = autogen.UserProxyAgent(
name="User_proxy",
system_message="Please describe the image for me.",
human_input_mode="TERMINATE",
max_consecutive_auto_reply=10,
code_execution_config={
"use_docker": False
},
)
BadRequestError: Context Length Exceeded for Multiple Images in Base64 Conversion
When attempting to process multiple images with the following code, an error occurs indicating that the context length exceeds the model's maximum token limit. The specific error is:
BadRequestError: Error code: 400 - {'error': {'message': "This model's maximum context length is 128000 tokens. However, your messages resulted in 144641 tokens. Please reduce the length of the messages.", 'type': 'invalid_request_error', 'param': 'messages', 'code': 'context_length_exceeded'}}Code:
when attempting to use multiple images, the following code results in an error: `import base64
Function to convert image file to base64
def convert_image_to_base64(image_path): with open(image_path, "rb") as image_file: return base64.b64encode(image_file.read()).decode("utf-8")
Data with image paths
data = [ { "title": "Step 1: Open the Application", "content": [ { "type": "image", "path": "step1_image.png" } ] }, { "title": "Step 2: Navigate to the Settings", "content": [ { "type": "image", "path": "step2_image.png" } ] } ]
Function to construct message with base64 images
def construct_message(data): message_parts = []
Construct the message with base64 image data
message = construct_message(data) `
The agents are as follows:
`agent1 = MultimodalConversableAgent( name="example-agent-1", max_consecutive_auto_reply=10, llm_config={"config_list": config_list_example, "temperature": 0.5, "max_tokens": 300}, system_message="You are an assistant who helps interpret and analyze image-based content.", ) agent2 = AssistantAgent( name="example-agent-2", max_consecutive_auto_reply=10, system_message="You are a support specialist who provides feedback on content interpretation.", )
user_proxy = autogen.UserProxyAgent( name="User_proxy", system_message="Please describe the image for me.", human_input_mode="TERMINATE", max_consecutive_auto_reply=10, code_execution_config={ "use_docker": False }, )
Set max_round to 5
groupchat = autogen.GroupChat(agents=[agent2, agent1], messages=[], max_round=5)
vision_capability = VisionCapability(lmm_config={"config_list": config_list_example, "temperature": 0.5, "max_tokens": 900}) group_chat_manager = autogen.GroupChatManager(groupchat=groupchat, llm_config=gpt4_llm_config) vision_capability.add_to_agent(group_chat_manager)
rst = user_proxy.initiate_chat( group_chat_manager, message=message, ) `