legate-boost tutorial #240
Conversation
|
Hi @RAMitchell, |
|
/ok to test d58f7a6 |
Signed-off-by: Rory Mitchell <[email protected]>
There was a problem hiding this comment.
I will fix the pre-commit errors, many of which are not related to this PR.
legate-boost.rst needs to be included in index.rst so sphinx knows to build it.
Have you been able to check the docs build locally? make html from the docs folder and then you can use a web server to preview the html pages.
Legate-boost currently has no dependency on legate-dataframe so I will have to add that for the tests to pass.
Generally speaking the examples look great. I don't think people will take seriously the comparison of 1 cpu core against a gpu for inference. I suspect most of the time here is latency inherent in legate and cupynumeric anyway - 1 gpu should be able to predict orders of magnitude faster than 1 cpu core, there just isn't enough work in this task to saturate the GPU. You can still keep all the timing code, I would just bump the cpus to 8 and then mention that this task is not compute bound. You will probably see a dramatic speed-up of GPU vs CPU during training with a neural network model if that is what you want to show.
|
/ok to test 370aff0 |
|
I think I've solved the other unrelated CI issues, and have removed the binary files. I added the new rst file to the index and built the docs locally and they look good! The tests that run the examples fail currently. You can run them locally like Please have a go at fixing these up. |
|
/ok to test b90ef34 |
Signed-off-by: Rory Mitchell <[email protected]>
|
/ok to test 0c0931e |
|
/ok to test a432096 |
|
/ok to test 07f19b2 |
|
Hi Rory, any update on this doc publishing? |
4 participants
legate-boost tutorial doc with code files