ChartMoE(ICLR2025 Oral)
Collection
[ICLR2025 Oral] ChartMoE: Mixture of Diversely Aligned Expert Connector for Chart Understanding
github: https://github.com/IDEA-FinAI/ChartMoE • 6 items • Updated
This is a reproduction of ChartMoE according to its official github repo, which has better performance on ChartQA(with/without PoT).
ChartMoE
ICLR2025 Oral
ChartMoE is a multimodal large language model with Mixture-of-Expert connector, based on InternLM-XComposer2 for advanced chart 1)understanding, 2)replot, 3)editing, 4)highlighting and 5)transformation.
Import from Transformers
To load the ChartMoE model using Transformers, use the following code:
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
ckpt_path = "IDEA-FinAI/chartmoe"
tokenizer = AutoTokenizer.from_pretrained(ckpt_path, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(ckpt_path, trust_remote_code=True).half().cuda().eval()
Quickstart & Gradio Demo
We provide a simple example and a gradio webui demo to show how to use ChartMoE. Please refer to https://github.com/IDEA-FinAI/ChartMoE.
Open Source License
The code is licensed under Apache-2.0.
- Downloads last month
- 12
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Install from pip and serve model
# Install vLLM from pip: pip install vllm# Start the vLLM server: vllm serve "Coobiw/ChartMoE_Reproduced"# Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Coobiw/ChartMoE_Reproduced", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'