imvladikon/Ling-3.0-tiny-GLM-5.3-Flash-surgery

Remove a __pycache__ entry that a local import left behind

Store weights in the released GLM-5.3-Flash layout

Add Ling and Flash chat template variants with Ling default

Replace weights with the tools/structured-output checkpoint

Set GLM-5.3-Flash base_model metadata as requested

Document experimental training use, generation limitations, and ten observed answers

Simplify verified default loading example and correct surgery model-card metadata

v0.4.0: minimize unused vision scaffold and verify native text outputs

v0.4.0: stock GLM save_pretrained export; no remote model code; verified text generation

v0.3.0: native MLA block distillation, preserved vocabulary and verified English generation

v0.2.0: activation-calibrated Ling to native GLM-5.3 text surgery

initial commit