Search results

[Ascend]: Qwen3多模态MOE架构模型使用EP训练报错

ascendeverything about Ascend supporteverything about Ascend support

Status: Open.

#643 In ByteDance-Seed/VeOmni;

· lxjhunan opened on Apr 10, 2026

[dist] bug: distribute_tensor uses src_data_rank=0 redundantly in load_model_weights when all ranks load from disk

Status: Open.

#637 In ByteDance-Seed/VeOmni;

· kahlun opened on Apr 9, 2026

TrainingArguments object has no attribute 'enable_mixed_precision'

bugSomething isn't workingSomething isn't working

Status: Open.

#636 In ByteDance-Seed/VeOmni;

· UserChen666 opened on Apr 9, 2026

Bring determinism to index over length on mapping dataset

Status: Open.

#630 In ByteDance-Seed/VeOmni;

· LiuzcEECS opened on Apr 2, 2026

The output instruct of VLM tranning is compressed to one or two.

Status: Open.

#578 In ByteDance-Seed/VeOmni;

· peikwang opened on Mar 17, 2026

Support UG Model: Janus, Bagel

Status: Open.

#534 In ByteDance-Seed/VeOmni;

· Coach257 opened on Mar 4, 2026

Support DiT Trainer

Status: Open.

#533 In ByteDance-Seed/VeOmni;

· Coach257 opened on Mar 4, 2026

RFC: Adding Support for GLM-5 Model

Status: Open.

#521 In ByteDance-Seed/VeOmni;

· Crystal-jiang opened on Feb 28, 2026

Potential BUG in training wan with SP

Status: Open.

#516 In ByteDance-Seed/VeOmni;

· zelingao98 opened on Feb 26, 2026

Multi-worker dyn batching

Status: Open.

#511 In ByteDance-Seed/VeOmni;

· Coach257 opened on Feb 25, 2026

train_flux.py在npu上训练，vit中的torch.nn.Conv2d算子运行时间特别长，导致flux训练无法正常进行

bugSomething isn't workingSomething isn't working

Status: Open.

#510 In ByteDance-Seed/VeOmni;

· yyyyrrrrif opened on Feb 25, 2026

[Feature Dev Track] Enhancing VeRL Integration

enhancementNew feature or requestNew feature or request

help wantedExtra attention is neededExtra attention is needed

hf_v5Related for transformers v5Related for transformers v5

road mapRoadmap for VeOmniRoadmap for VeOmni

Status: Open.

Feature

#506 In ByteDance-Seed/VeOmni;

· Luosuu opened on Feb 25, 2026