[WIP] fix loss oom (#1788)

Add host memory metrics to available_memory function (#1764)

fix(geo3k-vlm-sft): remove --apply-chat-template from SFT launch script (#1791)

fix qwen3.5 397B converting error when enable expert parallel (#1799)

pass critic role through to create RayTrainGroup (#1797)

[release] bump to v0.2.4 (#1777)

some minor cleanup

remove redundant envvar

update docs

(fix):not have encoder_only attr cause run failed (#1741)

[Fix]Fix some bugs/clean up (#1756)

fix typo (#1759)

Support FP8 conversion for Qwen3.5 (#1769)

[Fix] Fix duplicate Megatron LR scheduler resume when optimizer state is not loaded (#1775)

Add rollout trace timeline viewer (#1776)