SGD Optimized

AXL-Fixer-12M

Debug specialist. 20.9M params. PPL 1.52. Error-to-fix.

21M
Parameters
1.52
Perplexity
---
Training
24 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 on debug data. Self-debug cross-attention.
MetricValue
Final Loss0.4165
Perplexity1.52
Training Steps?
Training Time---

Usage

ollama create axl-fixer-12m -f Modelfile
ollama run axl-fixer-12m "def fibonacci():"
Generates minimal fixes from error traces. Pairs with AXL-Debugger.
FileSizeFormat
F16 GGUF24 MBFull precision
Q4_K_M GGUF24 MB4-bit quantized
GGUF files work with Ollama and llama.cpp. Q4_K_M is about 3x smaller than F16.
← All AXL Models