muon-model-test

BananaMind 2 Nano Muon FineWeb-Edu Test

This experimental base model uses the exact BananaMind 2 Nano architecture and tokenizer. It was trained from scratch with stock PyTorch Muon on the hidden matrices and AdamW on the tied embedding and normalization weights, using only streamed FineWeb-Edu data for 24,999,591,936 custom-tokenizer tokens.

Architecture

Training

The original Nano effective batch was 12 micro-batches x 11 accumulation steps = 132 sequences. This distributed run preserves that exact global batch. Ranks receive 16 or 17 sequences and scale their local mean losses so DDP's averaged gradient is the true 132-sequence global mean.

Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Banaxi-Tech/muon-model-test"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    trust_remote_code=True,
    dtype=torch.bfloat16,
    device_map="auto",
)

This is a base model, not an instruction-tuned chat model.