Janusz Lisiecki
Janusz Lisiecki
Thank you, @acecchini, for refreshing the topic. Currently, we are pursuing other priorities, but if you are willing to contribute to the project, we are happy to assist and guide...
I'm afraid that most, if not all, goes to the native part.
Hi @acecchini, Unfortunately, this enhancement is not as high on our ToDo list as you might wish. However, I encourage you to try implementing it on your own. It could...
@acecchini, Here are some thoughts on adding the functionality you're requesting: - Replace `kDaliDataloaderSeed` with a user-provided seed. To achieve this, for every operator that accepts the "shuffle_after_epoch" argument (like...
Hi @chenc29xpeng, Indeed, DALI returns not the output we may expect. Let us investigate this and get back to you soon.
Hi @chenc29xpeng, I'm sorry to say that there is no progress. The issue seems to be not related to DLAI itself but the GPU video decoding capability. I reached out...
@bpleshakov it could be but I cannot tell.
Hi @aafaqin, Thank you for reaching out. The current flow that employs the iterator is as follows: - Obtain a batch from the pipeline. If none has been scheduled, schedule...
Hi @aafaqin, If you want to run the inference the above configuration should work fine and you can skip the iterator and just use run method: ``` import torch o...
@aafaqin, are you observing this shift in the output of the DALI pipeline or from TensorRT (TRT)? Could you provide a self-contained code example that reproduces this issue? You can...