Lion Optimized

AXL-TestGen-15M

Test generation. 15.2M params. PPL 1.14. Context 2048 bytes.

15M
Parameters
1.14
Perplexity
10 min
Training
31 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 10MB test pairs. 173 steps in 10 min.
MetricValue
Final Loss0.1468
Perplexity1.14
Training Steps173
Training Time10 min

Usage

ollama create axl-testgen-15m -f Modelfile
ollama run axl-testgen-15m "def fibonacci():"
Generates unit tests from function signatures.
FileSizeFormat
F16 GGUF31 MBFull precision
Q4_K_M GGUF31 MB4-bit quantized
GGUF files work with Ollama and llama.cpp. Q4_K_M is about 3x smaller than F16.
← All AXL Models