Patryk Niemiec

EDIT: Didn't notice it's pjreedie repo. Opening issue on AlexeyAB fork got the same issue with 1080 ti, built with ZED 3.2, opencv 4.5 with CUDA. Training only for 1...

@Remco-Terwal-Bose see mentioned issue above or [here](https://github.com/AlexeyAB/darknet/issues/7153)

Thanks @eugen-ajechiloae-clearml, would help a lot. Keras introduced "Keras v3" format with `.keras` extension as recommended from TF2.13. Not sure if this is related to this issue, but would be...

Sure @eugen-ajechiloae-clearml . It's introduced in [2.13 release](https://github.com/tensorflow/tensorflow/releases/tag/v2.13.0) as default format in place of SavedModel. It can be created according to an example in [Keras documentation](https://keras.io/guides/serialization_and_saving/#saving). Here is an output...

Same for me with RTX 8000 and A6000 setup in MirrorStrategy with NCCL and Hierarchical CrossDeviceOps. I get a huge block of `tensorflow/core/framework/local_rendezvous.cc:405 Local rendezvous recv item cancelled.` before first...

> Hi @moberweger , > > I have replicated the reported behaviour with colab using TF v2.14, 2.15 and tf-nightly. Please find the [gist ](https://colab.research.google.com/gist/Venkat6871/534a54b6bfe499c8bd7b41fb148a02c6/62526_2-14_2-15-nightly.ipynb)here for reference. > > Thank...

Hello @Venkat6871 , @SuryanarayanaY . Is it possible to get some tensorflower assigned to help with this issue? I've managed to resolve this issue by using `tf.image.extract_patches` as slicing mechanism...

@jkhenning Im facing the same problem. We have docker group as `999` instead of `1000`, so I swapped guid with 999. I couldn't find docker user in /etc/passwd and `1000`...

> Wow this is open from almost a year ago... > > I think someone could get a lot citations / clicks if they did a proper benchmark of transformer...

Hello @DavisK-rgb . Have you resolved this bug?