T G_ F G O T_ Qwen2.5_7 B

Introduction

The TG_FGOT_QWEN2.5_7B model is a Russian-language, fine-tuned version based on the QWEN2.5-7B-Instruct model. It is specifically designed to generate Telegram posts in a style similar to the author from the official FGOT Telegram channel.

Architecture

• Base Model: Qwen/Qwen2.5-7B-Instruct
• Language: Russian
• Context Length: 4,096 tokens

Training

The model was fine-tuned using a dataset of 154 Telegram posts. The training process did not utilize a system prompt, allowing for a more flexible response style. The dataset used for fine-tuning is available here.

Guide: Running Locally

1. Clone the Repository: Download the model files from the model card.
2. Install Dependencies: Ensure that your environment includes the necessary libraries for running Hugging Face models, such as transformers and torch.
3. Load the Model: Use the transformers library to load the model and tokenizer.
4. Inference: Input a prompt, such as "Напиши пост про черепашек", to generate content.

For optimal performance, consider using cloud GPUs such as those available on AWS, Google Cloud, or Azure.

License

The model is released under the Apache-2.0 license, allowing for both commercial and non-commercial use.