LenguajeNatural.AI Chat e Instrucciones 2B

LenguajeNatural.AI/leniachat-gemma-2b-v0 is a 2.6 billion parameter model developed by LenguajeNatural.AI specifically for the Spanish-speaking community. It is based on google/gemma-2b and is distributed under the Apache 2.0 license. The model is exclusively trained in Spanish to maximize effectiveness for Hispanic users.

Key Capabilities

• Spanish-centric Training: Optimized for text generation, chat, and instruction understanding in Spanish.
• Multi-phase Training: Underwent three distinct training phases:
  • Multi-task learning using supervised datasets (FLAN-style).
  • High-quality instruction tuning for complex instructions.
  • Chat and abstractive QA training for fluid conversations.
• Context Length: Supports a maximum sequence length of 8192 tokens.

Use Cases

This model is designed for applications such as:

• Text generation in Spanish.
• Chatbots and virtual assistants for Spanish-speaking users.
• Resolving user queries and following instructions in conversational contexts.

Limitations

As a 2B parameter model, it shares inherent limitations common to models of its size. Users should evaluate its performance in specific contexts and be aware of potential biases or errors, despite efforts to minimize them during training.