x1 7 B v0.1

Introduction

The x1-7B-v0.1 is the inaugural large language model developed by BRAHMAI, designed with advanced reasoning capabilities. It is built on the Qwen2.5 7B architecture and fine-tuned using synthetic datasets derived from open-source models and custom pipelines. This model is particularly effective in handling reasoning-based questions, showcasing its thought process with <thinking> tags.

Architecture

The model is based on the Qwen2.5 7B architecture. It leverages the Transformers library and supports text generation tasks. The model is trained using a diverse set of datasets that enhance its reasoning and mathematical capabilities.

Training

The x1-7B-v0.1 model is fine-tuned on a synthetic dataset created from open-source models. It employs multiple datasets and custom agentic pipelines to enrich its reasoning skills. The model's training process includes simulating various reasoning scenarios to build its competence in handling complex queries.

Guide: Running Locally

1. Install Dependencies: Ensure you have Python and pip installed. Then, install transformers and torch libraries using pip.

   pip install transformers torch
   

2. Clone the Model Repository: Download the model files from the Hugging Face model hub.

   git clone https://huggingface.co/brahmairesearch/x1-7B-v0.1
   

3. Load the Model: Use the transformers library to load the model and tokenizer.

   from transformers import AutoTokenizer, AutoModelForCausalLM
   
   tokenizer = AutoTokenizer.from_pretrained("brahmairesearch/x1-7B-v0.1")
   model = AutoModelForCausalLM.from_pretrained("brahmairesearch/x1-7B-v0.1")
   

4. Run Inference: Input your text and obtain the model's reasoning-based response.

   inputs = tokenizer("What is larger? 9.24 or 9.8?", return_tensors="pt")
   outputs = model.generate(**inputs)
   print(tokenizer.decode(outputs[0]))
   

For optimal performance, using a cloud GPU service like AWS, Azure, or Google Cloud is recommended, as the model comprises 7 billion parameters.

License

The x1-7B-v0.1 model is licensed under the Apache-2.0 license, permitting use, modification, and distribution under specified terms.