anyreach-ai/dualturn-qwen2.5-mimi-0.5B

Update README.md

Update README.md

Add datasets, Training Data section, DualTurn Model & Code, Authors, updated Citation

Fix README: snapshot_download for ONNX, correct repo name

Fix ONNX: remove lm_head, output hidden_states [B,T,896] instead of logits [B,T,151936]

Merge config into modeling_dualturn.py — fixes trust_remote_code import

Remove separate configuration file

Fix: self-contained AutoModel with key remapping + proper head architecture

Remove old modeling.py

Add self-contained modeling.py + clean README with FP32 and ONNX examples

Remove turn_taking.py (replaced by self-contained modeling.py)

Add DualTurn-Qwen-Mimi: FP32 safetensors + ONNX KV cache

initial commit