|
|
--- |
|
|
library_name: transformers |
|
|
tags: |
|
|
- text-generation |
|
|
- cemp |
|
|
- assisted-living |
|
|
- LoRA |
|
|
- tinyllama |
|
|
--- |
|
|
|
|
|
# Model Card for `alfboss/cempbot-tiny` |
|
|
|
|
|
This model is a fine-tuned version of TinyLlama-1.1B-Chat-v1.0, specialized in answering questions related to CEMP (Comprehensive Emergency Management Plans) for Assisted Living Facilities in Florida. It has been trained using LoRA (Low-Rank Adaptation) with domain-specific instructions and responses generated by Evergreen Brain Pvt. Ltd. |
|
|
|
|
|
## Model Details |
|
|
|
|
|
### Model Description |
|
|
|
|
|
- **Developed by:** Jagdish Sharma and team at Evergreen Brain Pvt. Ltd. |
|
|
- **Shared by:** [@evergreen-brain](https://huggingface.co/evergreen-brain) |
|
|
- **Model type:** Causal Language Model (AutoModelForCausalLM) |
|
|
- **Language(s):** English |
|
|
- **License:** apache-2.0 |
|
|
- **Finetuned from model:** [TinyLlama/TinyLlama-1.1B-Chat-v1.0](https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0) |
|
|
|
|
|
### Model Sources |
|
|
|
|
|
- **Repository:** https://huggingface.co/alfboss/cempbot-tiny |
|
|
- **Demo:** Available via Hugging Face Inference Endpoints (e.g., `/generate` endpoint) |
|
|
|
|
|
## Uses |
|
|
|
|
|
### Direct Use |
|
|
|
|
|
This model can be used to: |
|
|
- Guide assisted living facilities in Florida on CEMP requirements |
|
|
- Answer domain-specific questions like document submission, emergency contact planning, AHCA compliance, etc. |
|
|
- Serve as an onboarding chatbot or compliance assistant for ALF administrators |
|
|
|
|
|
### Downstream Use |
|
|
|
|
|
This model can be integrated into: |
|
|
- Facility portals for onboarding staff |
|
|
- AI-powered chatbots |
|
|
- CEMP automation tools |
|
|
|
|
|
### Out-of-Scope Use |
|
|
|
|
|
- Not intended for generating general-purpose content |
|
|
- Not suitable for legal or emergency advice without expert review |
|
|
- Not trained for multi-turn conversation or open-domain chat |
|
|
|
|
|
## Bias, Risks, and Limitations |
|
|
|
|
|
- The model was fine-tuned on a narrow domain, so its responses may not generalize well to other topics |
|
|
- There may be hallucinations if prompted outside the CEMP compliance domain |
|
|
- Not suitable for critical compliance filings without human verification |
|
|
|
|
|
### Recommendations |
|
|
|
|
|
Always review model outputs with a compliance expert before submission. Use the model in supervised environments. |
|
|
|
|
|
## How to Get Started with the Model |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained("alfboss/cempbot-tiny") |
|
|
model = AutoModelForCausalLM.from_pretrained("alfboss/cempbot-tiny") |
|
|
|
|
|
prompt = "What is required in a Florida ALF CEMP plan?" |
|
|
inputs = tokenizer(prompt, return_tensors="pt") |
|
|
outputs = model.generate(**inputs, max_new_tokens=150) |
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |
|
|
|
