--- license: apache-2.0 library_name: peft tags: - generated_from_trainer base_model: BioMistral/BioMistral-7B model-index: - name: spanish_medica_llm results: [] --- # Model Card for SpanishMedicaLLM Más de 600 millones de personas hablantes del idioma español necesitan recursos, como los LLMs, para la obtención de información médica de forma libre y segura, cumpliendo con los objetivo del milenio: Salud y Bienestar, Educación y Calidad, Fin de la Pobreza propuestos por la ONU. Existen pocos LLM para el dominio médico en idioma español. El objetivo de este proyecto es crear un gran modelo de lenguaje (LLM; siglas en inglés) para el contexto médico en español permitiendo crear soluciones y servicios de información de salud en LATAM. El modelo contará con información de medicinas convencionales, naturales y tradicionales. Un resultado del proyecto es un conjunto de datos público del dominio médico que agrupa recursos de otras fuentes que permite crear o ajustar LLM. Los resultados del desempeño del LLM se comparan con otros modelos del state-of-the-art como BioMistral, Meditron, MedPalm. ## Model Details ### Model Description - **Developed by:** [Dionis López Ramos](https://www.linkedin.com/in/dionis-lopez-ramos/), [Alvaro Garcia Barragan](https://huggingface.co/Alvaro8gb), [Dylan Montoya](https://huggingface.co/dylanmontoya22), [Daniel Bermúdez](https://huggingface.co/Danielbrdz) - **Funded by:** SomosNLP, HuggingFace - **Model type:** Language model, instruction tuned - **Language(s):** Spanish (`es-ES`, `es-CL`) - **License:** apache-2.0 - **Fine-tuned from model:** [BioMistral/BioMistral-7B](https://huggingface.co/BioMistral/BioMistral-7B) - **Dataset used:** [somosnlp/SMC/](https://huggingface.co/datasets/somosnlp/SMC/) ### Model Sources - **Repository:** [spaces/somosnlp/SpanishMedicaLLM/](https://huggingface.co/spaces/somosnlp/SpanishMedicaLLM/tree/main) - **Paper:** "Comming soon!" - **Demo:** [spaces/somosnlp/SpanishMedicaLLM](https://huggingface.co/spaces/somosnlp/SpanishMedicaLLM) - **Video presentation:** [SpanishMedicaLLM | Proyecto Hackathon #SomosNLP ](https://www.youtube.com/watch?v=tVe_MC7Da6k) ## Uses ### Direct Use [More Information Needed] ### Out-of-Scope Use Los creadores del LLM no se hacen responsable de resultados nocivos que puedan generar. Se sugiere un proceso de evaluación riguroso con especialistas de los resultados generados. ## Bias, Risks, and Limitations [More Information Needed] ### Recommendations ## How to Get Started with the Model Use the code below to get started with the model. ``` [More Information Needed] ``` ## Training Details ### Training Data Dataset used was [somosnlp/SMC/](https://huggingface.co/datasets/somosnlp/SMC/) ### Training Procedure #### Training Hyperparameters **Training regime:** - learning_rate: 2.5e-05 - train_batch_size: 16 - eval_batch_size: 1 - seed: 42 - gradient_accumulation_steps: 4 - total_train_batch_size: 64 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 5 - training_steps: 2 - mixed_precision_training: Native AMP - ## Evaluation ### Testing Data, Factors & Metrics #### Testing Data Dataset used was a 20% from [somosnlp/SMC/](https://huggingface.co/datasets/somosnlp/SMC/) #### Factors [More Information Needed] #### Metrics [More Information Needed] ### Results [More Information Needed] ## Model Examination [optional] [More Information Needed] ## Environmental Impact Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700). - **Hardware Type:** GPU - **Hours used:** 4 Hours - **Cloud Provider:** [Hugginface](https://huggingface.co) - **Compute Region:** [More Information Needed] - **Carbon Emitted:** [More Information Needed] ### Model Architecture and Objective We used a LLM arquitecture to [BioMistral/BioMistral-7B](https://huggingface.co/BioMistral/BioMistral-7B) because if a foundational model trained in medical domain datasets. ### Compute Infrastructure [More Information Needed] #### Hardware Nvidia T4 Small 4 vCPU 15 GB RAM 16 GB VRAM #### Software - transformers==4.38.0 - torch>=2.1.1+cu113 - trl @ git+https://github.com/huggingface/trl - peft - wandb - accelerate - datasets [More Information Needed] ## License Apache License 2.0 ## Citation **BibTeX:** ``` @software{lopez2024spanishmedicallm, author = {Lopez Dionis, Garcia Alvaro, Montoya Dylan, Bermúdez Daniel}, title = {SpanishMedicaLLM}, month = February, year = 2024, url = {https://huggingface.co/datasets/HuggingFaceTB/cosmopedia} } ``` ## More Information This project was developed during the [Hackathon #Somos600M](https://somosnlp.org/hackathon) organized by SomosNLP. The model was trained using GPUs sponsored by HuggingFace. **Team:** - [Dionis López Ramos](https://huggingface.co/inoid) - [Alvaro Garcia Barragan](https://huggingface.co/Alvaro8gb) - [Dylan Montoya](https://huggingface.co/dylanmontoya22) - [Daniel Bermúdez](https://huggingface.co/Danielbrdz) ## Contact [optional] For any doubt or suggestion contact to: PhD Dionis López (inoid2007@gmail.com)