Hello. Thank your work. I wonder if we can enable DCA on other qwen3 models to support long context. I find the difference between config_1m.json and config.json is only the dual_chunk_attention_config field. So, can we enable this mechanism by adding this field?
