README.md

metadata

license: apache-2.0
datasets:
  - AI-MO/NuminaMath-TIR
language:
  - zho
  - eng
  - fra
  - spa
  - por
  - deu
  - ita
  - rus
  - jpn
  - kor
  - vie
  - tha
  - ara
metrics:
  - accuracy
base_model:
  - Qwen/Qwen2.5-0.5B-Instruct

NeuroCoder Qwen2.5-0.5B-Instruct-MemoryR

Overview

This is the Hugging Face checkpoint of Qwen2.5-0.5B-Instruct-MemoryR, a memory-augmented RL-tuned model based on Qwen2.5.

The model is introduced and analyzed in our paper: https://arxiv.org/abs/2504.02273

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("neurocoder/Qwen2.5-0.5B-Instruct-MemoryR")
model = AutoModelForCausalLM.from_pretrained("neurocoder/Qwen2.5-0.5B-Instruct-MemoryR")

# Example input
prompt = "What is the capital of France?"
inputs = tokenizer(prompt, return_tensors="pt")

# Generate output
outputs = model.generate(**inputs, max_new_tokens=50)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))