Fix RWTHDBIS inference device mismatch and add accelerate

[Kleinpenny]

Summary

• Fix RWTHDBIS inference device mismatch by moving inputs to the model’s actual device.
• Apply the same fix to both term typing and taxonomy discovery inference paths.
• Add missing accelerate dependency required by transformers.Trainer.
• Switch RWTHDBIS term typing and taxonomy discovery examples to use cuda by default.
• Ignore runtime outputs under examples/results/ and results/.

Background

Running RWTHDBIS examples on GPU triggered a runtime error during inference:
Expected all tensors to be on the same device because inputs were sent to self.device
while the model had already been moved to GPU by the trainer.

Changes

• ontolearner/learner/term_typing/rwthdbis.py: move inference inputs to model_device.
• ontolearner/learner/taxonomy_discovery/rwthdbis.py: same device alignment fix.
• Add accelerate>=0.26.0 to requirements.txt, pyproject.toml, and setup.py.
• examples/llm_learner_rwthdbis_term_typing.py: set device="cuda".
• examples/llm_learner_rwthdbis_taxonomy_discovery.py: set device="cuda".
• .gitignore: ignore examples/results/ and results/.

Impact

• Prevents CPU/GPU mismatch during inference in RWTHDBIS examples.
• Ensures required dependency is available for Trainer-based training.
• Examples default to GPU for faster training on CUDA machines.
• Avoids accidentally committing training artifacts.

Test plan

• python examples/llm_learner_rwthdbis_term_typing.py
• python examples/llm_learner_rwthdbis_taxonomy_discovery.py

[HamedBabaei]

Dear @Kleinpenny, thanks for your contribution. To proceed, can you update your branch with the latest version of the ontolearner/dev branch? As I see unwanted comments to metadata file, which is being automatically generated via our CI/CD pipeline, and it seems that there are a few changes on this PR.

Moreover, I can certenly understood why you have modified the .gitingore but this change is only for your local development, and it shouldn't be committed for PR. Additionally, I really appreciate your observation regarding the accelerate library; however, we intentionally left this up to the user since variant OS (like Mac OS) might not have support for such a library, which may cause an error in installation. So, I might as well request a revert on that.

Thank You.

I am looking forward to your next commits.

[Kleinpenny]

Dear Hamed,

Thanks for your suggestions and reviewing! I will edit this PR and follow your instructions to avoid conflict. I'm looking into the trainer of our approach inside, and i will make another PR soon. Hopefully i can back to you within this week.

[Kleinpenny]

Can be closed coz the changes has applied in PR #306