Search results

add sequence_parallel_size and error.

Status: Open.

#63 In Qihoo360/360-LLaMA-Factory;

· guotong1988 opened on Jul 30, 2025

使用多模态数据集微调Qwen2.5-VL，KeyError: 'sequence_parallel_attention'。问题出现在模型加载阶段，Transformers库中的Qwen2.5-VL模型不支持 sequence_parallel_attention 这种注意力实现

Status: Open.

#61 In Qihoo360/360-LLaMA-Factory;

· hybdxxw opened on Jul 21, 2025

src/data/template.py 出现未被定义的变量

Status: Open.

#60 In Qihoo360/360-LLaMA-Factory;

· diansheng opened on Jul 14, 2025

code和requirement.txt中对transformers的要求不一致

Status: Open.

#59 In Qihoo360/360-LLaMA-Factory;

· Huihao-Yang opened on Jun 5, 2025

使用流式加载报错，传递了额外参数：TypeError: IterableDataset.map() got an unexpected keyword argument 'num_proc'

Status: Open.

#57 In Qihoo360/360-LLaMA-Factory;

· CxsGhost opened on Jun 5, 2025

Support for Sequence Parallelism (SP) in the Multimodal Pipeline?

Status: Open.

#56 In Qihoo360/360-LLaMA-Factory;

· Eisenhower opened on May 29, 2025

对于sp， 当用ulysses替换ring_flash_attn时,为什么loss很大

Status: Open.

#55 In Qihoo360/360-LLaMA-Factory;

· cangshuli opened on May 13, 2025

关于华为NPU的支持

Status: Open.

#47 In Qihoo360/360-LLaMA-Factory;

· githubwqj opened on Apr 17, 2025

SP does NOT work with liger kernel

good first issueGood for newcomersGood for newcomers

Status: Open.

#36 In Qihoo360/360-LLaMA-Factory;

· XD-BDIV-NLP opened on Mar 22, 2025

360-LLAMA-Factory框架能否实现长序列的训练加速？还是说只是会减少显存占用？

Status: Open.

#30 In Qihoo360/360-LLaMA-Factory;

· StarDewXXX opened on Mar 15, 2025

Support for transformers versions < 4.46.1 will soon be discontinued due to issues with incorrect gradient accumulation.

Status: Open.

#29 In Qihoo360/360-LLaMA-Factory;

· chenzihan1 opened on Mar 14, 2025

When selecting data to pack, could an option be added to allow for random pack instead of using the default greedy pack?

Status: Open.

#27 In Qihoo360/360-LLaMA-Factory;

· cizhenshi opened on Mar 14, 2025