Rupesh Sapkota

The idea is to fine-tune existing LLMs (e.g., Gemma) or foundation models using approaches such as [Quantised Low-Rank Adaptation (QLoRA)](https://arxiv.org/abs/2305.14314) or a similar approach to tackle the link Prediction task....

ASWA callback broken for TP trainer or PL trainer with GPUs > 1

2

At the end of the each training epoch, when using ASWA callback, averaging operations are skipped by comparing the Global and Local Rank of current trainer, and hence no ASWA...

[Literal Embedding](https://github.com/dice-group/dice-embeddings/blob/122743f9823d80585cc30a8f6f6f53f7526d1e38/dicee/models/literal.py#L5) can be used to enhance entity embeddings in KGE using literals. [ReaLitE](https://link.springer.com/chapter/10.1007/978-3-031-94575-5_3) enhances relation embeddings in KGE using Literals. An approach that enhances both entity and relation embeddings...