RWKV-Red-Team/ARWKV-7B-Preview-0.1-NoG-32B

ARWKV🪿

Paper Link👁️ | Github✅

ARWKV-7B-FROM-32B (Preview 0.1)

Preview version with RWKV-7 time mixing and Transformer MLP

This version doesn't have the parameter "g". It freezes the MLP and uses 32B logits for distillation alignment.

📌 Overview

ALL YOU NEED IS RWKV

This is an early preview of our 7B parameter hybrid RNN-Transformer model, trained on 2k context length (only stage-2 applied, without SFT or DPO) through 3-stage knowledge distillation from Qwen2.5-7B-Instruct. While being a foundational version, it demonstrates:

• ✅ RWKV-7's efficient recurrence mechanism
• ✅ No self-attention, fully O(n)
• ✅ Constant VRAM usage
• ✅ Single-GPU trainability

Roadmap Notice: We will soon open-source different enhanced versions with:

• 🚀 16k+ context capability
• 🧮 Math-specific improvements
• 📚 RL enhanced reasoning model

How to use

pip3 install --upgrade rwkv-fla transformers

from transformers import AutoModelForCausalLM, AutoTokenizer


model = AutoModelForCausalLM.from_pretrained(
    "RWKV-Red-Team/ARWKV-7B-Preview-0.1-NoG-32B",
    device_map="auto",
    torch_dtype=torch.float16,
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(
    "RWKV-Red-Team/ARWKV-7B-Preview-0.1-NoG-32B"
)

🔑 Key Features

Component	Specification	Note
Architecture	RWKV-7 TimeMix + SwiGLU	Hybrid design
Context Window	2048 training CTX	Preview limitation
Training Tokens	40M	Distillation-focused
Precision	FP16 inference recommended(16G Vram required)	15%↑ vs BF16

🏗️ Architecture Highlights

Core Modification Flow

Qwen2.5 Decoder Layer:
- Grouped Query Attention
+ RWKV-7 Time Mixing (Eq.3)
- RoPE Positional Encoding
+ State Recurrence
= Hybrid Layer Output