Lion Optimized

AXL-Docs-8M

Documentation generation. 9.9M params. PPL 1.12. Context 2048 bytes.

10M
Parameters
1.12
Perplexity
10 min
Training
20 MB
GGUF
PropertyValue
ArchitectureMulti-Scale Transformer
d_model?
Attention Heads?
Layers per Scale?
Context Window2048 bytes
Downsample Factors[1, 2, 4]
Vocab Size258 (byte-level)
OptimizerLion
Retrained with Lion on 16MB documentation pairs. 215 steps in 10 min.
MetricValue
Final Loss0.1067
Perplexity1.12
Training Steps215
Training Time10 min

Usage

ollama create axl-docs-8m -f Modelfile
ollama run axl-docs-8m "def fibonacci():"
Adds docstrings with parameter descriptions.
FileSizeFormat
F16 GGUF20 MBFull precision
Q4_K_M GGUF20 MB4-bit quantized
GGUF files work with Ollama and llama.cpp. Q4_K_M is about 3x smaller than F16.
← All AXL Models