capyinn-gemma4-e2b-it-lora-v1-preview

CapyInn Gemma 4 E2B IT LoRA V1 Preview

Preview LoRA adapter for hospitality receptionist experiments.

This is an adapter-only release. It is not a standalone model. Load it with the base model:

google/gemma-4-E2B-it

Adapter

• Base model: google/gemma-4-E2B-it
• Adapter type: PEFT LoRA
• License: Apache 2.0
• Project: CapyInn hospitality receptionist dataset experiments

Limitations

This is a preview research adapter, not a production hotel receptionist system. It may produce incorrect, incomplete, or overconfident answers. Do not use it to make booking, payment, legal, safety, privacy, or operational decisions without a verified PMS/tool layer and human review.

Loading

Use the processor/tokenizer from the base model unless you intentionally need a local copy:

from peft import PeftModel
from transformers import AutoProcessor, AutoModelForCausalLM

base_model_id = "google/gemma-4-E2B-it"
adapter_id = "chuanman2707/capyinn-gemma4-e2b-it-lora-v1-preview"

processor = AutoProcessor.from_pretrained(base_model_id)
model = AutoModelForCausalLM.from_pretrained(base_model_id)
model = PeftModel.from_pretrained(model, adapter_id)

If the base model requires a different Transformers auto class in your environment, use the same class that successfully loads google/gemma-4-E2B-it.