Eva - AI Tutor and Teacher

Eva is an AI tutor and teacher designed to guide, educate, and help users learn effectively through clear explanations, examples, and interactive teaching methods.

Model Information

• Model Name: Eva
• Architecture: EvaForConditionalGeneration
• Base Type: Vision-Language Model
• Purpose: Educational tutoring and teaching assistance

Features

• Multi-modal capabilities (text, image, and video processing)
• Educational focus with tutoring methodology
• Interactive teaching approach
• Clear explanations and examples
• Adaptive learning support

Technical Specifications

• Model Type: eva
• Hidden Size: 4096
• Attention Heads: 32
• Hidden Layers: 36
• Max Position Embeddings: 262,144
• Vision Patch Size: 16x16
• Supported Formats: Text, Images, Videos

Usage

Eva is configured to act as a personal tutor, providing:

• Step-by-step explanations
• Educational guidance
• Interactive learning experiences
• Concept clarification
• Problem-solving assistance

Configuration Files

• config.json - Main model configuration
• chat_template.jinja - Chat template with Eva's teaching persona
• tokenizer_config.json - Tokenizer settings
• Model weights in safetensors format

Requirements

• Transformers library version 4.57.1 or compatible
• Support for bfloat16 precision
• Vision processing capabilities for multi-modal inputs