Vikhrmodels/Vikhr-Qwen-2.5-1.5B-Instruct

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kPublished:Oct 6, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Vikhr-Qwen-2.5-1.5B-Instruct is a 1.5 billion parameter instruction-tuned language model developed by Vikhrmodels, based on Qwen-2.5-1.5B-Instruct. It is specialized for high-efficiency text processing in both Russian and English, having been fine-tuned on the Russian-language GrandMaster-PRO-MAX dataset. This model excels at instruction generation, contextual responses, and text analysis, making it suitable for applications requiring precise bilingual text handling.

Loading preview...

Vikhr-Qwen-2.5-1.5B-Instruct Overview

Vikhr-Qwen-2.5-1.5B-Instruct is a 1.5 billion parameter instruction-tuned language model from Vikhrmodels, built upon the Qwen-2.5-1.5B-Instruct architecture. Its primary distinction is its bilingual specialization in Russian and English, achieved through supervised fine-tuning (SFT) on the proprietary Russian-language GrandMaster-PRO-MAX dataset.

Key Capabilities

  • Bilingual Proficiency: Optimized for high-efficiency text processing and understanding in both Russian and English.
  • Instruction Following: Designed to generate precise responses based on given instructions.
  • Contextual Responses: Capable of providing coherent and contextually relevant answers.
  • Text Analysis: Excels in analyzing and interpreting textual data.
  • Training Methodology: Fine-tuned using a synthetic dataset of 150k instructions, incorporating Chain-Of-Thought (CoT) and GPT-4-turbo prompts to enhance accuracy and coherence.

Good For

  • Applications requiring robust performance in Russian language tasks.
  • Use cases demanding accurate instruction generation and contextual understanding in a bilingual (RU/EN) setting.
  • Integration into user-facing applications and services where efficient text processing is crucial.

Quantized variants (GGUF, MLX 4-bit, MLX 8-bit) are also available for optimized deployment.