Summary
Split from #332 — add Baichuan causal LM model support only.
Tasks
- Add Baichuan model config adapter (
csrc/models/baichuan/)
- Register
"baichuan" in config_factory.cpp classic_models list
- Register
"baichuan" in python/infinilm/auto_config.py
- Add Baichuan weight remapping (
W_pack → q/k/v_proj) in modeling_utils.py
- Update
examples/jiuge.py for Baichuan tokenization and chat prompt handling
Parent issue: #332
Summary
Split from #332 — add Baichuan causal LM model support only.
Tasks
csrc/models/baichuan/)"baichuan"inconfig_factory.cppclassic_models list"baichuan"inpython/infinilm/auto_config.pyW_pack→q/k/v_proj) inmodeling_utils.pyexamples/jiuge.pyfor Baichuan tokenization and chat prompt handlingParent issue: #332