Add generation_config.json (eos_token_id=7) for Transformers.js

Set kv_cache_dtype + torch_dtype to float32 (matches actual ONNX weights)

Drop dangling model.onnx ref from use_external_data_format

Patch ONNX layout + config for Transformers.js (onnx/model_q4)

Upload config.json

Upload tokenizer_config.json

Upload decoder ONNX model (int4) via onnxruntime-genai

initial commit