Rombos-LLM-V2.6-Qwen-14b

Introduction

Rombos-LLM-V2.6-Qwen-14b is an advanced text generation model developed as an upgrade to the previous version, Rombos-LLM-V2.5-Qwen-14b. This model leverages state-of-the-art techniques to improve performance across various benchmarks.

Architecture

The model is based on the Qwen2.5-14B-Instruct, utilizing the Transformers library. It is designed for text generation tasks and is compatible with safetensors for efficient storage.

Training

The model has been evaluated on multiple datasets including:

• IFEval (0-Shot): Achieved a strict accuracy of 52.14.
• BBH (3-Shot): Achieved a normalized accuracy of 49.22.
• MATH Lvl 5 (4-Shot): Achieved an exact match score of 28.85.
• GPQA (0-shot): Achieved a normalized accuracy of 17.
• MuSR (0-shot): Achieved a normalized accuracy of 19.26.
• MMLU-PRO (5-shot): Achieved an accuracy of 48.85.

Continuous finetuning methods are applied to improve model performance.

Guide: Running Locally

To run the Rombos-LLM-V2.6-Qwen-14b model locally, follow these steps:

1. Installation: Ensure you have Python and the Transformers library installed. You can install Transformers via pip:

   pip install transformers
   

2. Model Download: Download the model weights from Hugging Face:

   from transformers import AutoModelForCausalLM, AutoTokenizer
   
   tokenizer = AutoTokenizer.from_pretrained("rombodawg/Rombos-LLM-V2.6-Qwen-14b")
   model = AutoModelForCausalLM.from_pretrained("rombodawg/Rombos-LLM-V2.6-Qwen-14b")
   

3. Inference: Use the model for text generation:

   inputs = tokenizer("Your input text here", return_tensors="pt")
   outputs = model.generate(**inputs)
   print(tokenizer.decode(outputs[0], skip_special_tokens=True))
   

4. Hardware Requirements: For optimal performance, it is recommended to use cloud GPUs. Providers like AWS, Google Cloud, or Paperspace offer suitable GPU instances.

License

The Rombos-LLM-V2.6-Qwen-14b model is licensed under the Apache-2.0 License, allowing for extensive use and modification.