SGD Optimized

AXL-Debugger-8M

Bug fixing. 14.1M params. PPL 1.49. Self-debug module.

14M
Parameters
1.49
Perplexity
---
Training
15 MB
GGUF
PropertyValue
ArchitectureMulti-Scale Transformer
d_model?
Attention Heads?
Layers per Scale?
Context Window256 bytes
Downsample Factors[1, 2, 4]
Vocab Size258 (byte-level)
OptimizerSGD
Trained with SGD for 10 min. Self-debug module cross-attends error messages to source code.
MetricValue
Final Loss0.3979
Perplexity1.49
Training Steps?
Training Time---

Usage

ollama create axl-debugger-8m -f Modelfile
ollama run axl-debugger-8m "def fibonacci():"
Processes error messages and generates code fixes.
FileSizeFormat
F16 GGUF15 MBFull precision
Q4_K_M GGUF15 MB4-bit quantized
GGUF files work with Ollama and llama.cpp. Q4_K_M is about 3x smaller than F16.
← All AXL Models