/Dataset: Contains datasets for 6 time reasoning tasks, unified into a classification task format. Due to size limitations, RCW and ECG datasets are not included in this repository but will be made available in the final version through alternative means./SupervisedModels: Contains supervised time series models used for comparison./LLMs: Contains LLM-based methods, including traditional numerical modeling and the proposed VL-Time method.
Install the required libraries by running:
pip3 install -r ENV.txt- 
Navigate to the SupervisedModels directory:
cd /SupervisedModels - 
Run experiments using scripts in the
scriptsfolder. For example:bash Transformer.sh
- Use 
--root_pathto specify the dataset - Use 
--modelto specify the model 
 - Use 
 
- 
Navigate to the LLMs directory:
cd /LLMs - 
Set up API key:
- Add your API key in 
/LLMs/Method/LMM.py - Specifically, set the 
openai_api_key 
 - Add your API key in 
 - 
Prepare visualization data:
- Place the visualization data in the 
LLMs/Datasetfolder - Currently, data for visual modeling is provided
 
 - Place the visualization data in the 
 - 
Run experiments:
bash Scripts.sh
- Set 
modal="L"for traditional numerical modeling - Set 
modal="V"for VL-Time visual modeling - Use 
--modelto specify different LLMs - Set 
--num_shot_per_class=0for zero-shot reasoning; use values >0 for few-shot setting - Use 
--hint="Please solve this problem step by step"for chain-of-thought reasoning 
 - Set 
 
This library is constructed based on the following repos: