Align chat template `enable_thinking` default with generation_config.json and the model card

fix(rope): YaRN attention_factor is (0.1*ln(factor)+1)*attn_factor, not the bare multiplier

Use chat_template.jinja as the single source: drop the {% include %} chat_template field from tokenizer_config.json

Update README.md (#8)

Update README.md (#7)

small self-contained fixes

initial commit