Pinned issues
Issues
- 4
- 1
微调训练视频数据读取问题
#30 opened by wangyin717 - 3
二阶段微调训练的问题
#29 opened by wangyin717 - 11
DeepSpeed的PP需要相同的seq-length(collate时注意padding)和batchsize(将dataloader的`drop_last`设为True)
#25 opened by Youngluc - 0
logo upload
#28 opened by Coobiw - 1
video-chat example 2(language: en)
#27 opened by Coobiw - 3
请教一下关于SFT的问题
#26 opened by df2046df - 10
AI 无法理解图片
#24 opened by liuling19941216 - 2
请问如何减少训练时长
#14 opened by xiyuanhao - 2
请问是否支持 流水线并行 推理
#15 opened by valencebond - 1
Probably lower loss when use `train_pipeline.py`
#22 opened by Coobiw - 0
video-chat example upload
#23 opened by Coobiw - 1
- 3
是否支持QWEN-14B的INT4的量化版本?
#18 opened by yumianhuli2 - 3
请问如何用Qwen-14B进行重新训练
#11 opened by delltower - 6
关于知乎中提到的多模态接入方案问题
#17 opened by cszhengyh - 5
special token
#7 opened by PangziZhang523 - 6
训练loss异常
#5 opened by balabala2023 - 1
Code confusion
#2 opened by abbhay - 2
请问Qwen-7B的权重文件是只需要LFS的吗?还是全部文件都要呢?
#16 opened by cszhengyh - 3
- 1
和千问VL做过比较吗?
#6 opened by FoolishMao - 4
deepspeed training, meet the error "ValueError: optimizer got an empty parameter list"
#13 opened by sunnnnnnnny - 1
Error:safetensors_rust.SafetensorError: Error while deserializing header: HeaderTooLarge
#10 opened by ccccai239 - 1
huggingface 下载的Qwen7B-chat/None
#9 opened by molyswu - 1
- 3
学习率一直是1e-4不会下降?
#4 opened by Minami-su - 4
ValueError: unknown url type: '/export/dataset/minigpt4/minigpt4_minigpt4qwen_format.json'
#3 opened by Minami-su