config/models.yaml

# Model Configuration for Vision Language Models and Language Models
# This file contains model configurations for easy integration

models:
  # InternVL Vision-Language Models
  InternVL3-8B:
    name: "InternVL3-8B"
    model_id: "OpenGVLab/InternVL3-8B"
    model_type: "internvl"
    description: "Fastest model, good for quick processing"
    supported_quantizations:
      - "non-quantized(fp16)"
      - "quantized(8bit)"
    default_quantization: "non-quantized(fp16)"
    
  InternVL3-14B:
    name: "InternVL3-14B"
    model_id: "OpenGVLab/InternVL3-14B"
    model_type: "internvl"
    description: "Balanced performance and quality"
    supported_quantizations:
      - "non-quantized(fp16)"
      - "quantized(8bit)"
    default_quantization: "quantized(8bit)"
    
  InternVL3-38B:
    name: "InternVL3-38B"
    model_id: "OpenGVLab/InternVL3-38B"
    model_type: "internvl"
    description: "Highest quality, requires significant GPU memory"
    supported_quantizations:
      - "non-quantized(fp16)"
      - "quantized(8bit)"
    default_quantization: "quantized(8bit)"
    
  InternVL3_5-8B:
    name: "InternVL3_5-8B"
    model_id: "OpenGVLab/InternVL3_5-8B"
    model_type: "internvl"
    description: "Fastest model, good for quick processing"
    supported_quantizations:
      - "non-quantized(fp16)"
      - "quantized(8bit)"
    default_quantization: "non-quantized(fp16)"
    
  # Qwen Language Models (Text-only)
  Qwen2.5-7B-Instruct:
    name: "Qwen2.5-7B-Instruct"
    model_id: "Qwen/Qwen2.5-7B-Instruct"
    model_type: "qwen"
    description: "Qwen2.5 7B instruction-tuned model for text generation"
    supported_quantizations:
      - "non-quantized(fp16)"
      - "quantized(8bit)"
    default_quantization: "quantized(8bit)"
    
  Qwen2.5-14B-Instruct:
    name: "Qwen2.5-14B-Instruct"
    model_id: "Qwen/Qwen2.5-14B-Instruct"
    model_type: "qwen"
    description: "Qwen2.5 14B instruction-tuned model for better text generation"
    supported_quantizations:
      - "non-quantized(fp16)"
      - "quantized(8bit)"
    default_quantization: "quantized(8bit)"

# Default model selection
default_model: "InternVL3-8B"