v14 README

v14: format-normalized thinking

docs: add Bilibili link

v13 README

v13: switchable thinking (A+B toggle, difficulty-graded CoT)

v12 README

v12: fix GGUF thinking degradation (thinking ratio 36.7%)

v10: final balanced version

v6: fix code bugs + context-aware comfort

v5: multilingual projects + scene-aware length + emoji

v4: casual conversation incremental finetune (60 short dialogs)

v3: multilingual (zh/en/yue) + 3500 code instructions, lr 1e-4

docs: add GGUF links

docs: note poor general-conversation quality

v2: fix template overfitting, lr 1e-4

Upload README.md with huggingface_hub

LycheeAI-coder-1.7b: code-focused LLM fine-tuned from Qwen3-1.7B (QLoRA)

initial commit