Esbieta-EC-Qwen-2.5-3B

Model Description

Esbieta-EC-Qwen-2.5-3B is a fine-tuned version of Qwen-2.5-3B specifically trained on Ecuadorean cuisine and recipes from the Esbieta YouTube channel. This model specializes in understanding and generating content related to Ecuadorean food, cooking techniques, and traditional recipes, with a particular focus on Esbieta's unique culinary style and adaptations.

Model Details

  • Base Model: Qwen-2.5-3B
  • Model Type: Language Model for Recipe Generation and Food Knowledge
  • Language: Spanish (Ecuadorean dialect)
  • License: [Same as Qwen-2.5-3B]
  • Developed by: Pablo Carrera, Robin Kroonen
  • Dataset: Esbieta Dataset
  • Size: 3B parameters

Intended Use

This model is designed for:

  • Generating and understanding Ecuadorean recipes
  • Answering questions about Ecuadorean cuisine
  • Providing cooking instructions in the style of Esbieta
  • Understanding local Ecuadorean cooking terms and ingredients

Perfect for food enthusiasts, home cooks, and anyone interested in Ecuadorean cuisine or following Esbieta's cooking style.

Training Data

The model was trained on the pabloce/esbieta dataset, which includes:

  • Transcribed recipes from Esbieta's YouTube channel
  • Traditional Ecuadorean recipes
  • Cooking instructions and techniques
  • Ingredient lists and substitutions common in Ecuador

How to Use

Installation

pip install transformers

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained("pabloce/esbieta-ec-qwen-2.5-3B")
tokenizer = AutoTokenizer.from_pretrained("pabloce/esbieta-ec-qwen-2.5-3B")

# Example: Generate a recipe
prompt = "Dame la receta del encebollado ecuatoriano:"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=500)
recipe = tokenizer.decode(outputs[0])

Example Prompts

  1. Recipe Generation:
"¿Cómo se prepara el locro de papa estilo Esbieta?"
"Dame la receta del ceviche ecuatoriano"
  1. Ingredient Questions:
"¿Qué ingredientes necesito para hacer guatita?"
"¿Cuál es el sustituto del ají peruano en Ecuador?"
  1. Cooking Techniques:
"¿Cómo se hace el refrito base ecuatoriano?"
"Explica la técnica de Esbieta para hacer hornado"

Limitations

  • The model's knowledge is limited to Ecuadorean cuisine and Esbieta's recipes
  • May not be familiar with very regional or uncommon variations of dishes
  • Best performance in Spanish, particularly Ecuadorean Spanish
  • May need guidance with exact measurements and cooking times

Training Details

  • Training Data: Esbieta YouTube transcriptions and Ecuadorean recipe corpus
  • Fine-tuning Approach: [Specific training details]
  • Validation: Tested against known Esbieta recipes and traditional Ecuadorean dishes

Citations

If you use this model, please cite:

@misc{esbieta-ec-qwen,
  author = {Pablo Carrera},
  title = {esbieta-ec-qwen-2.5-3B: An Ecuadorean Cuisine Language Model},
  year = {2025},
  publisher = {HuggingFace},
  journal = {HuggingFace Model Hub},
  howpublished = {\url{https://huggingface.co/pabloce/esbieta-ec-qwen-2.5-3B}}
}

Acknowledgments

  • Esbieta YouTube Channel for the original content and recipes
  • Qwen team for the base model
  • Ecuadorean cooking community for traditional recipes and techniques

Contact

License

MIT

Downloads last month
34
Safetensors
Model size
3.09B params
Tensor type
BF16
·
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and HF Inference API was unable to determine this model's library.

Datasets used to train pabloce/esbieta-ec-qwen-2.5-3B

Spaces using pabloce/esbieta-ec-qwen-2.5-3B 3