Kazuki1450/Qwen3-1.7B-Base_csum_3_10_tok_python_1p0_0p0_1p0_grpo_42_rule

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:2BQuant:BF16Ctx Length:32kPublished:Mar 18, 2026Architecture:Transformer Warm

Kazuki1450/Qwen3-1.7B-Base_csum_3_10_tok_python_1p0_0p0_1p0_grpo_42_rule is a 2 billion parameter language model, fine-tuned from Qwen/Qwen3-1.7B-Base using the TRL framework. This model incorporates the GRPO training method, as detailed in the DeepSeekMath paper, which is designed to enhance mathematical reasoning capabilities. With a context length of 32768 tokens, it is optimized for tasks requiring robust logical and mathematical processing.

Loading preview...

Model Overview

This model, Kazuki1450/Qwen3-1.7B-Base_csum_3_10_tok_python_1p0_0p0_1p0_grpo_42_rule, is a fine-tuned variant of the Qwen/Qwen3-1.7B-Base architecture. It leverages the TRL framework for its training process.

Key Training Methodology

A significant aspect of this model's development is the application of GRPO (Gradient Regularized Policy Optimization). This method, introduced in the paper "DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models", aims to improve the model's mathematical reasoning abilities. The training procedure was tracked and can be visualized via Weights & Biases.

Technical Specifications

  • Base Model: Qwen3-1.7B-Base
  • Training Framework: TRL (version 0.29.0)
  • Core Training Method: GRPO
  • Context Length: 32768 tokens

Potential Use Cases

Given its foundation in Qwen3-1.7B-Base and the specialized GRPO training, this model is likely well-suited for:

  • Tasks requiring enhanced mathematical reasoning.
  • Applications benefiting from a robust base model with fine-tuned logical capabilities.
  • Exploration of models trained with advanced reinforcement learning techniques like GRPO.