Property	Value
Developed by	CubicLabs
Architecture	Multi-Scale Transformer
Parameters	677056
Optimizer	Lion
Attention	SDPA
Vocab Size	258 (byte-level)
Context Window	256 bytes
d_model	64
Attention Heads	4
Layers per Scale	2
Downsample Factors	[1, 2, 4]
License	Apache 2.0

Sources

Repository: GitHub
Organization: CubicLabs

Uses

Direct Use

Demo/testing model (Shakespeare).

python

import torch
from multiscale_transformer.model.model import MultiScaleTransformer
from multiscale_transformer.training.tokenizer import ByteTokenizer
ckpt = torch.load("axl_micro_600k.pt", map_location="cpu")
model = MultiScaleTransformer(config)
model.load_state_dict(ckpt["model_state_dict"])
model.eval()
tokenizer = ByteTokenizer()
ids = torch.tensor([tokenizer.encode("def hello():")], dtype=torch.long)
with torch.no_grad():
    out = model.generate(ids, max_new_tokens=50, temperature=0.8)
print(tokenizer.decode(out[0].tolist()))

Out-of-Scope Use

Not for production code generation. Not for code generation tasks. For integration with tools like Continue.dev, LlamaIndex, or LangChain, use the Python API server which provides OpenAI-compatible endpoints.

Bias, Risks, and Limitations

Byte-level perplexity is not comparable to BPE-level perplexity. Shakespeare-trained demo model. Not for code generation. Note: GGUF files for Ollama use a simplified single-stack encoder. For full AXL quality, use the Python API server.

Recommendations

Use for prototyping and experimentation, not production code generation.
Byte-level perplexity (258 vocab) is not comparable to BPE-level perplexity (32K vocab).
For better results, use the Lion-optimized version if available.

Training Details

Training Data

Retrained with Lion on Shakespeare. 2435 steps in 2 min. PPL 1.04.

Preprocessing

Byte-level tokenization with vocabulary size 258 (256 bytes + BOS + EOS). No vocabulary training required.

Speeds, Sizes, Times

Metric	Value
Training Steps	2435
Training Time	2 min
Final Loss	0.0747

Evaluation

Metrics

Perplexity on held-out Python code using byte-level tokenization.

Results

Metric	Value
Perplexity (byte-level)	1.04
Final Loss	0.0747
Training Steps	2435
Training Time	2 min

Summary: Demo model for testing architecture. Shakespeare-trained.

Environmental Impact

Property	Value
Hardware	AMD Ryzen 5 5600G
Hours Used	0.033
Carbon Emitted	0.0014 kg CO2
Cloud Provider	None (local CPU)

Technical Specifications

Model Architecture

Multi-Scale Transformer with three parallel encoder stacks at resolution scales 1x, 2x, and 4x. Cross-scale attention connects all scale pairs. Adaptive gating fusion. SwiGLU feed-forward. RoPE positional encoding.

Compute Infrastructure

Property	Value
Hardware	AMD Ryzen 5 5600G (6 cores, 12 threads)
RAM	16 GB
GPU	None (CPU-only)

Citation

bibtex

@misc{axl_2026,
  title={AXL: AXL-Micro-600K - Multi-Scale Transformer for CPU Code Generation},
  author={Cubic},
  year={2026},
  url={[https://huggingface.co/CubicLabs](https://huggingface.co/CubicLabs)}
}

How to Get Started

With Ollama

bash

ollama create axl-micro-600k -f Modelfile
ollama run axl-micro-600k "def fibonacci():"

With Python