-
Final sync: pristine config.json with LogicL1ForCausalLM and zero .py files
-
Pristine release: pure logic_l1 architecture with zero .py files
-
AROM Labs: Establish completely independent native LogicL1 architecture implementation
-
Align config.json dimensions to Qwen2-1.5B checkpoint (hidden_size=1536)
-
Set architecture to LogicL1ForCausalLM with pristine file tree
-
Native logic_l1 registration: purge all .py files for a clean standard repository structure
-
Clean repository file tree: remove custom .py wrappers and use native qwen2 backbone
-
Complete repository repair: restore custom model wrappers, config, and stable chat template
-
Register model natively as logic_l1 using absolute-import wrapper classes
-
Upload folder using huggingface_hub