SPIRAL OctoThinker-8B Multi-Agent Model

This model was trained using the SPIRAL (Self-Play Iterative Reinforcement learning for Adaptation and Learning) framework.

Model Details

Base Model: OctoAI/OctoThinker-8B
Training Framework: SPIRAL
Checkpoint: step_00320
Model Size: 8B parameters
Training Date: 2025-08-31

Training Configuration

The model was trained with self-play on multiple environments:

KuhnPoker-v1
TicTacToe-v0
SimpleNegotiation-v1

Training Parameters

{
  "learning_rate": "1e-6",
  "train_batch_size": 128,
  "num_ppo_epochs": 2,
  "temperature": 1.0,
  "max_model_len": 16384,
  "environments": [
    "KuhnPoker-v1",
    "TicTacToe-v0",
    "SimpleNegotiation-v1"
  ],
  "base_model": "OctoThinker-8B",
  "framework": "SPIRAL"
}

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

tokenizer = AutoTokenizer.from_pretrained("the-acorn-ai/spiral-octothinker-8b-multi-three-games-step00320")
model = AutoModelForCausalLM.from_pretrained(
    "the-acorn-ai/spiral-octothinker-8b-multi-three-games-step00320",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

# Generate text
inputs = tokenizer("Your prompt here", return_tensors="pt")
outputs = model.generate(**inputs, max_length=100)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)

License

This model is licensed under the Apache License 2.0.