Edwardmark

how the model perform on object detection?

5

how the model perform on object detection?

Questions about DSP in OpenDiT.

1

Hi, thanks for your great work. Do you apply the DSP in OpenDiT in Open-Sora-V1.1 training and Inference?

不使用DDP，只用lmdb，速度很慢，比原始imread还慢

4

``` def folder2lmdb(anno_file, name="train", write_frequency=5000, num_workers=16): ids = [] annotation = [] for line in open(anno_file,'r'): filename = line.strip().split()[0] ids.append(filename) annotation.append(line.strip().split()[1:]) lmdb_path = osp.join("app_%s.lmdb" % name) isdir = os.path.isdir(lmdb_path) print("Generate...

虹膜识别讨论

2

你好 愿意一起讨论下虹膜识别吗？我的邮箱[email protected]

for masked video generation, what is the meaning of edit ratio?

1

for masked video generation, what is the meaning of edit ratio?

why T5 must use fp32 instead of fp16 or bf16?

support for open-sora 1.1?

1

hi, do openDit support for open-sora 1.1?

Thanks for your great work. Will you open-source the test code of running Megatron-SP and DeepSpeed-Ulyssess in the DSP paper? The DSP figure 1, the shape before attention all gather...

questions about the paper DSP figure 1.

1

Hi, thanks for the great work, I have some questions about your DSP paper. how to do residual add before attention and after attention, the shape is different for megatron...

should nproc-per-node equals to sequence_parallel_size?

3

Great work. I have a question, should nproc-per-node equals to sequence_parallel_size when do open-sora inference using DSP? Thanks in advance.