Jurgis Pods

Hi @debasishg. I'm back from vacation and willing to support porting to Kafka 2.2. Please let me know about the current state of work.

@debasishg I've added an email address to the commits (if this is still needed).

It took me some time to get the CLA approved, which is now done. @debasishg Shall we get on with the issue? Do you have a branch/PR for the work...

We are hitting a similar problem: We use a custom partitioner to partition by 1) a custom payload field and 2) by time (year, month, day), like in this one:...

Having a similar problem with `TemporalFusionTransformer` and `QuantileLoss`. I can see that `QuantileLoss._device` is set to `cuda:0` even though the TFT model was mapped to CPU using ``` TemporalFusionTransformer.load_from_checkpoint(best_model_path, map_location='cpu')...

No. I'm not sure why, but in the meantime, not even `TemporalFusionTransformer.load_from_checkpoint(best_model_path)` works. It might be because I upgraded to Pytorch-Lightning 2.0.2. It now throws an error when calling `_load_from_checkpoint`...

it might be worth mentioning there are these two warning in the log when trying to load a model from a checkpoint: ``` UserWarning: Attribute 'loss' is an instance of...

I provided a PR for this issue: #1374, feedback welcome!

Try replacing the line ``` RUN sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys F42ED6FBAB17C654 ``` with ``` RUN sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu $(lsb_release -sc) main" > /etc/apt/sources.list.d/ros-latest.list' >/dev/null 2>&1...