PWZER
I think the copy-on-read is not suitable for larger variables, For larger variables, the performance of using shared locks will be better？
> @PWZER, I tried to execute the given code on tensorflow nightly version and found that Keras training speed is 37ms/step. It is similar to the Estimator training speed. Kindly...
> @PWZER, We can see that the tf-nightly version you are using v1.12.1, whereas if you check the gist attached, on tf-nightly version Keras training speed is 37ms/step. It is...
It is same problem with you, but the version of supervisor is last (3.3.3), and the Python version is 2.7.5. So. How should I do?
I got it. but I recommend that should force a partition when create a new topic, not use the default partition num configuration?
This is a good solution, we want to solve this problem too, When can it be merged?
My problem has been solved. It was due to insufficient GPU memory. When I changed from FP32 to FP16, the problem no longer occurred. I am curious why it is...
@tropxy Can you try this? - Remove `openapi` function - Update `quart_api_doc` like ```python quart_api_doc( app, config=app.__schema__, url_prefix="/api/doc/", title="API doc", editor=True ) ```
> Hi, any updates on this? @tropxy Do you have a demo to reproduce the problem?