yibinlei/effir-mistral-drop-8-mlp

TEXT GENERATIONConcurrent Unit Cost:1Model Size:7BQuant:FP8Context Size:4kTool Calling:SupportedPublished:Jun 26, 2026Architecture:Transformer Featherless Exclusive Cold

The yibinlei/effir-mistral-drop-8-mlp is a 7 billion parameter Mistral-based dense retriever model developed by yibinlei. This model incorporates direct layer dropping, enhancing its efficiency for retrieval tasks. It is specifically designed for applications requiring efficient information retrieval using a Mistral architecture.

Loading preview...

Model Overview

The yibinlei/effir-mistral-drop-8-mlp is a 7 billion parameter model based on the Mistral architecture, developed by yibinlei. It is characterized as a dense retriever that utilizes a technique called "direct layer dropping" to optimize its performance.

Key Capabilities

  • Efficient Retrieval: The model is designed for efficient information retrieval tasks, leveraging its dense retriever architecture.
  • Layer Dropping: Incorporates direct layer dropping, which can contribute to improved efficiency and potentially reduced computational overhead during inference.
  • Mistral Base: Built upon the Mistral foundation, suggesting a strong base for language understanding and generation capabilities, adapted for retrieval.

Usage Notes

  • Requires trust_remote_code=True for proper loading and functionality.
  • The provided code snippet demonstrates how to load the model, tokenizer, and configuration, including handling input embeddings and merging PEFT adapters.