Michał Zientkiewicz

Thanks for the reply. Unfortunately, the question is no longer relevant, since I've moved to a different (in-house) framework. I needed slightly different functionality - specifically, much more native support...

Hello @benchd, Please check your device id. You said you can access "device 0", but your DALI snippet specifies device 1. ```Python pipe = my_pipe(batch_size=1, num_threads=1, device_id=1) ^^^^^^^^^^^ ```

Hello @jackdaw213, If your loop isn't too long, then it should be possible to unroll it to the maximum length. With that you could use conditional execution to emulate shorter...

> Hello @mzient, thank you for your response > > > If your loop isn't too long, then it should be possible to unroll it to the maximum length. With...

@jackdaw213 I'm quite sure there are some constructs that work in "regular" mode but not in "debug". Can you try to run your code without debugging?

@jackdaw213 Thank you for checking non-debug pipeline. Currently there's no way to go from GPU to CPU within a single pipeline. We're actively working on relaxing the execution model to...

@jackdaw213 Hello again, GPU->CPU copy is now possible as an opt-in feature (it requires using a special executor, enabled with `exec_dynamic=True` in your `pipeline_def`). With that you can call `.cpu()`...

@jackdaw213 Thank you for reporting the issue. This is indeed a bug that'll be fixed in the December release of DALI. Meanwhile, as a workaround just pass the result of...

The original mirror is back and has been ever since. No need to switch.

Yeah, I'd recommend playing with the seed a bit to get an example with three different sequence lengths, preferably without a single-image sequence. --- View entire conversation on ReviewNB