Model Card of germanInstructionBERTcased for Bertology

A minimalistic german instruction model with an already good analyzed and pretrained encoder like dbmdz/bert-base-german-cased. So we can research the Bertology with instruction-tuned models, look at the attention and investigate what happens to BERT embeddings during fine-tuning.

The training code is released at the instructionBERT repository. We used the Huggingface API for warm-starting BertGeneration with Encoder-Decoder-Models for this purpose.

Training parameters

• base model: "dbmdz/bert-base-german-cased"
• trained for 3 epochs
• batch size of 16
• 40000 warm-up steps
• learning rate of 0.0001

Purpose of germanInstructionBERTcased

InstructionMBERT is intended for research purposes. The model-generated text should be treated as a starting point rather than a definitive solution for potential use cases. Users should be cautious when employing these models in their applications.