chargoddard/Yi-34B-Llama

chargoddard/Yi-34B-Llama Overview

This model is a re-implementation of the 01-ai/Yi-34B architecture, specifically adapted to be compatible with standard Llama modeling code. This adaptation allows for easier integration into existing Llama-based workflows and tools.

Key Characteristics

• Architecture: Based on the Yi-34B model from 01-ai, with internal tensor renaming for Llama compatibility.
• Parameters: Features 34 billion parameters, providing significant capacity for various natural language processing tasks.
• Context Length: Supports a substantial context window of 32768 tokens, enabling the processing of longer inputs and generating more coherent, extended outputs.
• Loading: The model itself can be loaded without requiring the trust_remote_code flag, simplifying deployment. However, the associated tokenizer may still require it, or users can opt for the llama-tokenizer branch for a Llama tokenizer class.

Good For

• Developers and researchers familiar with the Llama ecosystem looking for a high-parameter model.
• Applications requiring a large context window for detailed understanding or generation.
• Experimentation with the Yi-34B architecture within a Llama-compatible framework.