ValiantLabs/CodeLlama-13B-Fireplace

TEXT GENERATIONPricing:Input $1.5 / Output $2.1Concurrent Unit Cost:1Model Size:13BQuant:FP8Context Size:4kPublished:Jan 18, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

ValiantLabs/CodeLlama-13B-Fireplace is a 13 billion parameter model built on the Llama 2 architecture, specifically fine-tuned from CodeLlama-13b-Instruct-hf. It specializes in function calling and code instruction, leveraging the glaive-function-calling-v2 dataset for its training. This model is designed to excel in scenarios requiring structured function calls and code-related chat interactions, offering improved output structure for a better user experience.

Loading preview...

ValiantLabs/CodeLlama-13B-Fireplace Overview

Fireplace-13b is a specialized 13 billion parameter language model developed by Valiant Labs, built upon the Llama 2 architecture. It uses CodeLlama-13b-Instruct-hf as its base and is specifically fine-tuned for enhanced function calling and code instruction capabilities. This model is part of Valiant Labs' "Build Tools" initiative, aiming to provide useful open-source functionalities.

Key Capabilities

  • Function Calling: Optimized to interpret user requests and generate structured function calls, with a recommended format for system messages and function delivery.
  • Code Instruction/Chat: Excels at understanding and responding to code-related queries and instructions, leveraging the base CodeLlama's inherent coding strengths.
  • Improved Output Structure: Version 1.1 features enhancements to the output format for function calls, indicated by <functioncall> and <|endoftext|> delimiters.
  • General Instruct-Chat: While specialized, it also retains general instruction-chat capabilities.

Good for

  • Developers building applications that require precise function invocation from natural language.
  • Tasks involving code generation, explanation, or debugging through conversational interfaces.
  • Integrating AI assistants into development workflows where structured tool use is critical.