Lee Yang

Unfortunately, Java API support in TF has been spotty with [deprecation warnings](https://www.tensorflow.org/api_docs/java/org/tensorflow/package-summary) and [no API stability guarantees](https://www.tensorflow.org/jvm/install). We initially tried to support Java when the API was updated regularly with...

I don't see anything obvious from your logs. Given that it looks like the evaluator process stalled/quit, I'd check for CPU and memory usage on that node (when it's running)...

If you already have a trained model (and it fits in memory), then the simplest way to run inferencing in a Spark job is to use something like [this example](https://github.com/yahoo/TensorFlowOnSpark/blob/master/examples/mnist/keras/mnist_inference.py)....

@jiqiujia assuming that your model won't change over the course of the job, you can just cache the model in the python worker processes via a global variable. Just check...

@jahidhasanlinix Not quite sure what you're doing here... *.pt are PyTorch models. Have you converted a TensorFlow model to PyTorch (or vice versa)?

@jahidhasanlinix Unfortunately, I think that code looks like it's beyond the scope of what TFoS is trying to do. Decima presumably integrates with (or replaces) the spark scheduler itself, while...

My guess is that your `model_dir` needs to fully-specify the HDFS path, e.g. `hdfs://default/...` Note that the example uses spark-local mode (for simplicity).

> Pls also fix the linter failure: https://github.com/leewyang/spark/actions/runs/3397174449/jobs/5649073867#step:16:71 Updated to latest master, which got rid of the linter error, but it added a new "appveyor" check, which seems to be...

BTW, I'm seeing a change in behavior in the `pandas_udf` when used with `limit` in the latest master branch of spark (vs. 3.3.1), per this example code: ``` import numpy...

@WeichenXu123 Yes, using `df.limit(10).cache().withColumn` makes it only process 10 rows inside the pandas_udf, which addresses the performance issue, thanks!