--- language: - en license: apache-2.0 model-index: - name: AllyArc/llama_allyar results: - task: type: text-generation name: Text Generation dataset: name: chat_imitate type: AllyArc/chat_imitate split: test metrics: - type: bleu value: 0.5 name: BLEU - type: confusion_matrix value: 0.5 name: Confusion Matrix - type: glue value: 0.5 name: GLUE - type: mse value: 0.5 name: MSE - type: squad value: 0.5 name: SQUAD - type: wiki_split value: 0.8 name: Wiki Split --- # Model Card for AllyArc This model card describes AllyArc, an educational chatbot designed to support autistic students with personalized learning experiences. AllyArc uses a fine-tuned Large Language Model to interact with users and provide educational content. ## Model Details ### Model Description AllyArc is an innovative chatbot tailored for the educational support of autistic students. It leverages a fine-tuned LLM to provide interactive learning experiences, emotional support, and a platform for students to engage in conversational learning. - **Developed by:** Zainab, a computer science student and MLH Top 50. - **Model type:** Conversational Large Language Model - **Language(s) (NLP):** Primarily English, with potential multilingual support. - **Finetuned from model:** Mistral 7b. - [Dataset generation Script](https://colab.research.google.com/drive/1zqmu6vQn1Rb0OJPAoRBBYcI7qw0H_LCH#scrollTo=xOj-BuKARJrH) - [llama Finetuning Script](https://colab.research.google.com/drive/1dz49DEzCKiE2103A3Kdb8BLIed_aCFi_?usp=sharing) ## Uses ### Direct Use AllyArc can be directly interacted with by students and educators through a conversational interface, providing instant responses to queries and aiding in learning. ### Downstream Use The model can be integrated into educational platforms or applications as a support tool for autistic students, offering personalized assistance. ### Out-of-Scope Use AllyArc is not designed for high-stakes decisions, medical advice, or any context outside of educational support. ## Bias, Risks, and Limitations While designed to be inclusive, there is a risk of unintended bias in responses due to the training data. The model may not fully understand or appropriately respond to all nuances of human emotion and communication. ### Recommendations Educators should monitor interactions and provide regular feedback to improve AllyArc's accuracy and sensitivity. Users should be aware of the model's limitations and not rely on it for critical decisions. ## How to Get Started with the Model on Google Colab To explore and interact with AllyArc using Google Colab: 1. Open the [AllyArc Interactive Colab Notebook](https://colab.research.google.com/drive/1MiGTw7nKMFbE8FllpVAW66DTmQSzOTFd?usp=sharing). 2. Go to `File > Save a copy in Drive` to create a personal copy of the notebook. 3. Obtain a Hugging Face API token by creating an account or logging in at [Hugging Face](https://huggingface.co/). 4. In your copied notebook, replace `YOUR_HUGGING_FACE_TOKEN_HERE` with your actual Hugging Face token. 5. Follow the instructions in the notebook to install necessary libraries and dependencies. 6. Run the cells step by step to initialize and interact with the AllyArc model. Please ensure you have the appropriate permissions and quotas on Google Colab to run the model without interruption. ## How to Get Started with the AllyArc Model Locally To run the AllyArc model on your local machine, follow these steps: 1. Ensure you have Python installed on your system. 2. Install the necessary Python packages by running: ```bash pip install transformers tokenizers sentencepiece ``` 3. Obtain a Hugging Face API token by creating an account or logging in at [Hugging Face](https://huggingface.co/settings/tokens). 4. Set an environment variable for your Hugging Face token. You can do this by running the following command in your terminal (replace `` with your actual token): ```bash export HUGGING_FACE_API_KEY= ``` 5. Create a new Python script or open a Python interactive shell and input the following code: ```python import os from huggingface_hub import hf_hub_download from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline # Replace with your Hugging Face token HUGGING_FACE_API_KEY = os.environ.get("HUGGING_FACE_API_KEY") model_id = "ZainabF/allyarc_finetune_model_sample" filenames = [ "pytorch_model.bin", "added_tokens.json", "config.json", "generation_config.json", "special_tokens_map.json", "spiece.model", "tokenizer_config.json", "pytorch_model.bin.index.json" ] for filename in filenames: downloaded_model_path = hf_hub_download( repo_id=model_id, filename=filename, token=HUGGING_FACE_API_KEY ) print(f"Downloaded {filename} to {downloaded_model_path}") # Initialize the tokenizer and model tokenizer = AutoTokenizer.from_pretrained(model_id, legacy=False) model = AutoModelForSeq2SeqLM.from_pretrained(model_id) # Set up the pipeline for text generation text_gen_pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer, max_length=1000) # Generate a response response = text_gen_pipeline("How I'm upset that I got low mark at math, please help me") print(response) ``` 6. Execute the script to download the model and interact with it. Please ensure that your environment variables are correctly set, and that the necessary packages are installed before running the script. The script will download the model files and then initialize the model for text generation, allowing you to input prompts and receive responses.