Skip to content

Latest commit

 

History

History
159 lines (134 loc) · 5.14 KB

model_data_doc.md

File metadata and controls

159 lines (134 loc) · 5.14 KB

model_data.py API Reference

Related Modules

This module provides data structures used throughout the DeerTick project. For an overview of how it fits into the larger system, see the DeerTick Main Documentation.

Overview

This script loads model data from a CSV file and voice sample data from another CSV file, organizing the information into easily accessible dictionaries.

Global Variables

model_data

  • Type: dict
  • Description: A nested dictionary containing all data for each model.
  • Structure: {model_name: {column_name: value}}
  • Fields:
    • id: str - Unique identifier for the model
    • name: str - Name of the model
    • created: int - Timestamp of when the model was created
    • description: str - Detailed description of the model
    • context_length: int - Maximum context length the model can handle
    • per_request_limits: str or None - Limits per request, if any
    • architecture.modality: str - The modality of the model (e.g., "text->text", "text+image->text")
    • architecture.tokenizer: str - The tokenizer used by the model
    • architecture.instruct_type: str - The instruction type for the model
    • pricing.prompt: float - Price per token for prompts
    • pricing.completion: float - Price per token for completions
    • pricing.image: float - Price for image processing (if applicable)
    • pricing.request: float - Price per request (if applicable)
    • top_provider.context_length: int or None - Context length for the top provider
    • top_provider.max_completion_tokens: int or None - Maximum completion tokens for the top provider
    • top_provider.is_moderated: bool - Whether the top provider is moderated
    • preferred_provider: str - The preferred provider for this model
    • model_type: str - Type of the model (e.g., "llm")
  • Example usage: model_data['Meta: Llama 3.2 3B Instruct']['context_length']

model_index

  • Type: dict
  • Description: Maps model names to their integer index in the CSV file.
  • Structure: {model_name: index}
  • Example usage: model_index['Meta: Llama 3.2 3B Instruct']

index_to_model

  • Type: dict
  • Description: Maps integer indices to model names.
  • Structure: {index: model_name}
  • Example usage: index_to_model[0]

voice_samples

  • Type: dict
  • Description: Contains voice sample URLs for different voice names.
  • Structure: {voice_name: url}
  • Example usage: voice_samples['voice_name']

Schema

classDiagram
    class ModelData {
        +Dict[str, Dict[str, Any]] model_data
        +Dict[str, int] model_index
        +Dict[int, str] index_to_model
        +Dict[str, str] voice_samples
    }

    class Model {
        +str id
        +str name
        +int created
        +str description
        +int context_length
        +str per_request_limits
        +Architecture architecture
        +Pricing pricing
        +TopProvider top_provider
        +str preferred_provider
        +str model_type
    }

    class Architecture {
        +str modality
        +str tokenizer
        +str instruct_type
    }

    class Pricing {
        +float prompt
        +float completion
        +float image
        +float request
    }

    class TopProvider {
        +int context_length
        +int max_completion_tokens
        +bool is_moderated
    }

    class VoiceSample {
        +str voice_name
        +str url
    }

    ModelData "1" -- "*" Model : contains
    Model "1" -- "1" Architecture : has
    Model "1" -- "1" Pricing : has
    Model "1" -- "1" TopProvider : has
    ModelData "1" -- "*" VoiceSample : contains
Loading

Functions

This script does not define any functions. All operations are performed at the module level when the script is run.

Usage Examples

  1. Get the context length of a specific model:

    context_length = model_data['Meta: Llama 3.2 3B Instruct']['context_length']

  2. Get the index of a specific model:

    index = model_index['Meta: Llama 3.2 3B Instruct']

  3. Get the model name for a specific index:

    model_name = index_to_model[0]

  4. Get the URL for a specific voice sample:

    url = voice_samples['voice_name']

  5. Iterate through all models and print their pricing:

    for model_name, model_info in model_data.items():
    print(f"Model: {model_name}")
    print(f"  Prompt price: {model_info['pricing.prompt']}")
    print(f"  Completion price: {model_info['pricing.completion']}")

  6. Get the description of a model:

    description = model_data['Meta: Llama 3.2 3B Instruct']['description']

  7. Check if a model supports image processing:

    supports_image = model_data['Meta: Llama 3.2 3B Instruct']['architecture.modality'] == 'text+image->text'

Notes

  • The script assumes the existence of 'models.csv' and 'samples.csv' files in the same directory.
  • Any changes to the structure of these CSV files may require modifications to the script.
  • The script prints the number of loaded models and voice samples for verification purposes.
  • Some fields may be empty or None for certain models, so it's good practice to check for their existence before using them.