Janusz Lisiecki
Janusz Lisiecki
I think I identified the problem and https://github.com/NVIDIA/DALI/pull/4723 is an attempt to fix it. Please check [the nightly build](https://docs.nvidia.com/deeplearning/dali/user-guide/docs/installation.html#pip-nightly-and-weekly-releases) that follows the merge of the fix.
Hi @elmuz, > It seems to me that the frame number is returned as a GPU tensor, which if I understand correctly cannot be used as an index over other...
Hi @byungsoo-oh, You can just use [the `slice` operator](https://docs.nvidia.com/deeplearning/dali/user-guide/docs/operations/nvidia.dali.fn.slice.html), but you decode the sample fully first and then trim it (or pad using [`pad` operator](https://docs.nvidia.com/deeplearning/dali/user-guide/docs/operations/nvidia.dali.fn.pad.html)). Still, it sounds like a...
@byungsoo-oh thanks! We will look into it.
DALI 1.22 has been released with this functionality included.
Closing as stalled.
Closing as stalled.
Hi @thomas-bouvier, Please use one file with the whole data set description. The reader will shard it later.
Hi @thomas-bouvier , I'm happy that the file reader works as expected. In the case of the external source, you need to make sure that in each epoch all samples...
Hi @thomas-bouvier, If I understand correctly: ``` inf = self.data_set_len * shard_id // num_shards sup = self.data_set_len * (shard_id + 1) // num_shards self.files = np.array(self.files[inf:sup]) self.labels = np.array(self.labels[inf:sup]) ```...