ByteDance-Seed/VeOmni v0.1.4

v0.1.4

Repository: ByteDance-Seed/VeOmni

Tag: v0.1.4

Published: 2025-12-07T01:44:57Z

Prerelease: yes

Release notes:

Highlights

• We now have GPU and NPU in CI and enable some tests on them. We call for more contribution of tests if you are interested!
• Now VeOmni model/dataset/dataloader/checkpointer/chat_template/preprocess are registry-based, making adding new of them and customization more easily
• Update pyproject.toml for uv-based env management. We now only suggest you to install veomni through uv.

What's Changed

• update README.md by @Fazziekey in https://github.com/ByteDance-Seed/VeOmni/pull/188
• [dist] fix: refactor fsdp2 grad norm clipping by @Luosuu in https://github.com/ByteDance-Seed/VeOmni/pull/185
• [misc] fix: reset hf init flag for random init by @Luosuu in https://github.com/ByteDance-Seed/VeOmni/pull/176
• Fix Qwen3-Moe MFU by @zhihaofang1017 in https://github.com/ByteDance-Seed/VeOmni/pull/125
• [model] fix:avoid cpu-device sync for qwenvl on npu by @wey-code in https://github.com/ByteDance-Seed/VeOmni/pull/190
• [task] fix: replace DataArguments with MyDataArguments and remove duplicated step2token saving by @MuyaoLi-jimo in https://github.com/ByteDance-Seed/VeOmni/pull/189
• [ci] test: CI env test by @FoolPlayer in https://github.com/ByteDance-Seed/VeOmni/pull/201
• [ckpt][fix]release cuda mem after dcp sync save by @EricOlivier in https://github.com/ByteDance-Seed/VeOmni/pull/207
• [misc] feat: update uv support for aarch platform for Ascend+Kunpeng … by @pjgao in https://github.com/ByteDance-Seed/VeOmni/pull/148
• [data] fix: fix exception raised when fetching current_device on NPU by @ji-huazhong in https://github.com/ByteDance-Seed/VeOmni/pull/211
• [ci] test: fix data_ci by @Coach257 in https://github.com/ByteDance-Seed/VeOmni/pull/222
• [ci] test: add npu ci env by @FoolPlayer in https://github.com/ByteDance-Seed/VeOmni/pull/219
• title: [data] feat: Implement extensible data preprocessor registry by @TimYangst in https://github.com/ByteDance-Seed/VeOmni/pull/203
• [ci]Add NPU support to data and model test by @Crystal-jiang in https://github.com/ByteDance-Seed/VeOmni/pull/224
• [ci]Add Ascend NPU native support to the unit test code by @Crystal-jiang in https://github.com/ByteDance-Seed/VeOmni/pull/208
• [ci] chore: add gemini config & test by @FoolPlayer in https://github.com/ByteDance-Seed/VeOmni/pull/229
• [test] ci: add device api check for tests by @onehaitao in https://github.com/ByteDance-Seed/VeOmni/pull/213
• [core] feat: registry for dataset & dataloader & checkpointer & ckpt_to_state_dict & chat_template & preprocess by @Coach257 in https://github.com/ByteDance-Seed/VeOmni/pull/230
• [dist] fix: make OptimizerState EP-dim aware to fix its dcp saving by @Luosuu in https://github.com/ByteDance-Seed/VeOmni/pull/228
• Automatically add the "ascend" label by @Crystal-jiang in https://github.com/ByteDance-Seed/VeOmni/pull/234
• [helper]:fix npu profiling by @Feng0w0 in https://github.com/ByteDance-Seed/VeOmni/pull/214
• helper: degrade veomni_patch functions to warnings/no-op by @iqiancheng in https://github.com/ByteDance-Seed/VeOmni/pull/197
• [ci] fix: dataloader in e2e ckpt test by @Luosuu in https://github.com/ByteDance-Seed/VeOmni/pull/233
• [feat] nccl_timeout by @brook-cpp in https://github.com/ByteDance-Seed/VeOmni/pull/217
• Automatically apply "ascend" label to issues and PRs by @Crystal-jiang in https://github.com/ByteDance-Seed/VeOmni/pull/239
• chore: Upgrade PyTorch dependencies to 2.8.0 and flash-attention to 2.8.3 by @TimYangst in https://github.com/ByteDance-Seed/VeOmni/pull/242
• [version] update transformers version to 4.57.0 by @phdddd in https://github.com/ByteDance-Seed/VeOmni/pull/243
• feat: distributed checkpointer support customized backend by @Ziyi-Wang in https://github.com/ByteDance-Seed/VeOmni/pull/182
• [ckpt] refactor: remove unused output_dir parameter from ckpt_to_state_dict by @TimYangst in https://github.com/ByteDance-Seed/VeOmni/pull/248
• [config, omni, dis] fix: quick fix for sft of Wan2.1-I2V-14B-480P by @zbian99 in https://github.com/ByteDance-Seed/VeOmni/pull/240
• [model] fix: Update @check_model_inputs decorator for transformers 4.57+ compatibility by @TimYangst in https://github.com/ByteDance-Seed/VeOmni/pull/252
• [core] fix is_x_backend by @brook-cpp in https://github.com/ByteDance-Seed/VeOmni/pull/251
• [data] fix: quick fix for exception raised when building dit dataloader on NPU by @zbian99 in https://github.com/ByteDance-Seed/VeOmni/pull/246
• upgrade: Upgrade transformers from v4.57.0 to v4.57.3 by @yiwzhao in https://github.com/ByteDance-Seed/VeOmni/pull/249
• [core] feat: model registry by @Coach257 in https://github.com/ByteDance-Seed/VeOmni/pull/258
• [dist] feat: unified veomni grad norm clipping by @Luosuu in https://github.com/ByteDance-Seed/VeOmni/pull/205
• [task]fix: fix train.sh NPROC_PER_NODE calculation logic on the NPU by @Crystal-jiang in https://github.com/ByteDance-Seed/VeOmni/pull/227
• [chore]: cache ep group by @heidongxianhua in https://github.com/ByteDance-Seed/VeOmni/pull/231

New Contributors

• @zhihaofang1017 made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/125
• @MuyaoLi-jimo made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/189
• @FoolPlayer made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/201
• @EricOlivier made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/207
• @pjgao made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/148
• @ji-huazhong made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/211
• @TimYangst made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/203
• @Crystal-jiang made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/224
• @onehaitao made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/213
• @Feng0w0 made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/214
• @iqiancheng made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/197
• @brook-cpp made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/217
• @phdddd made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/243
• @zbian99 made their first contribution in https://github.com/ByteDance-Seed/VeOmni/pull/240
• @yiwzhao made their...