Amey9766/phi2-fine-tuned-hotel-concierge-merged

pipeline_tag: text-generation base_model: microsoft/phi-2 tags: - phi-2 - lora-merged - text-generation - hotel-concierge license: other # phi-2 has its own license; see Links below

Phi-2 Hotel Concierge — Fine-Tuned (Merged)

Author/Maintainer: @Amey9766
Base model: microsoft/phi-2 (Microsoft)
Summary: This repository provides merged weights of a phi-2 model fine-tuned via (Q)LoRA on a hotel-concierge style dataset for helpful, concise Q&A.

Attribution & License

• The underlying model is microsoft/phi-2 (© Microsoft) and is subject to its license/terms.
• This repo contains fine-tuned weights produced by @Amey9766. Any additional license you apply here covers only your fine-tuned deltas/merged weights and does not replace or override the base model license.
• Users must comply with both the base model’s license and any license for these weights.

Intended Use

• Hotel-concierge style assistance: check-in/out, amenities, local tips, transport, etc.
• Not for medical, legal, or safety-critical advice.

Limitations & Risks

• May hallucinate or produce outdated information.
• No browsing or live data.
• English-centric; other languages may degrade.

Training (high-level)

• Method: (Q)LoRA → merged with base using merge_and_unload.
• Data: curated hotel-concierge Q&A prompts (brief notes).
• Example hyperparams: (epochs/steps/lr etc., if you want to disclose).

Inference

from huggingface_hub import InferenceClient
client = InferenceClient("Amey9766/phi2-fine-tuned-hotel-concierge-merged")
prompt = "<|system|>: You are a hotel concierge.\n<|user|>: Do you have late checkout?\n<|assistant|>:"
print(client.text_generation(prompt, max_new_tokens=200, temperature=0.7))

Acknowledgements

• Microsoft for microsoft/phi-2.
• TRL / PEFT / Transformers for fine-tuning tools.

Links

• Base model: https://huggingface.co/microsoft/phi-2
• Base model license/terms: see model card above.