bertin-project/Gromenauer-7B

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:May 17, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Gromenauer-7B by bertin-project is a 7 billion parameter Spanish language model built on the Mistral architecture. It is specifically designed to understand and generate high-quality Spanish text, having been trained on an extensive literary corpus. This model excels at capturing linguistic nuances, styles, and contexts found in Spanish literature, making it ideal for Spanish-centric natural language processing tasks.

Loading preview...

Overview

bertin-project's Gromenauer-7B is a 7 billion parameter language model based on the Mistral architecture, specifically engineered for the Spanish language. It has been trained on an extensive literary corpus to ensure high-quality text generation and a deep understanding of Spanish linguistic nuances, styles, and contexts.

Key Capabilities

  • High-Quality Spanish Text Generation: Optimized for producing fluent and contextually appropriate Spanish text.
  • Linguistic Nuance Capture: Training on a broad literary corpus allows it to grasp subtle linguistic variations.
  • Mistral Architecture Benefits: Leverages the efficiency and performance characteristics of the Mistral model family, including features like Flash Attention and Sliding Window Attention.

Technical Specifications

  • Model Type: Mistral
  • Sequence Length: 8192 tokens
  • Hidden Dimension: 4096
  • Number of Layers: 32
  • Attention Heads: 32 (8 Key-Value Heads)
  • Sliding Window Attention: 4096 tokens

Training Details

  • Tokenizer: Uses the mistralai/Mistral-7B-v0.1 tokenizer.
  • Learning Rate: 1e-5 with Adam optimizer and Cosine schedule.
  • Training Steps: 7000 steps.

Good For

  • Applications requiring robust Spanish language understanding and generation.
  • Tasks involving literary analysis or style transfer in Spanish.
  • Developers seeking a specialized, high-performance Spanish LLM.