More Guidance Needed on Training Models on Own Datasets about neuralkg HOT 3 CLOSED

Thanks very much for your many useful suggestions and I respond to your suggestions as follows.

1. would it be possible to provide more details to guide users in training different models on their datasets, as well as details of the config file parameters, and the command line to run each pipeline?
   Actually, we are in the process of improving our documentation to include the parameters explanation of the config file, and we will soon be posting a blog on our website to guide users in training models on their datasets, and recording animated gifs to visualize them. For each pipeline, we will also shortly be providing both scripts and config files to reproduce the results of models
2. Questions about the documentation of the BaseDataModule.train_dataloader example for processing image data.
   This part is PyTorchLightning's description, thank you for pointing it out, we will change it as soon as possible to avoid any misunderstanding. The dataset is given as triples (h, r, t). If indexed dictionaries of entities and relations are provided, the index will be read automatically, otherwise, it will be created automatically, the dataset structure can see FB15K237, and the dataset does not need to be processed manually, just select the appropriate pipeline and NeuraKG will preprocess the dataset accordingly.

from neuralkg.

Sounds great! Thank you so much for taking the time to share these details. Looking forward to the updates!

from neuralkg.

We have updated what was to be done in the last response.

• We have released the script files and configuration files used to reproduce the model results
• We have written notebooks on colab to guide users through the use of our tools and posted blogs on our website to present detailed examples of using neuralkg. We also show animated gifs in README to show the training and testing process.
• In addition, we have updated the documentation with some basic parameter descriptions.

For more updates please see our news, this issue will be closed.

from neuralkg.