Ethan Yanjia Li

thanks for the advice. i'll add it when i have time. `valid_anchors_wh` is the `width, height` tuple for anchors.

i'm thinking if you have a `tfrecords` directory in the same level of this `tfrecords_mpii.py` script. basically, what this script does is to take annotations and images from one location...

so the script actually create tfrecords in a directory called `tfrecords` i made a fix here: https://github.com/ethanyanjiali/deep-vision/pull/10 or you can manually create it like: ``` dataset |-- tfrecords_mpii.py |--tfrecords --|--train_0055_of_0064.tfrecords...

As a data point, I tried to replicate the model in TF2 and got similar # of parameters like reported above. This is for MNext 1.0 so make divisible shouldn't...

what was the problemetic version of transformers?

Good idea but it also makes this project less intuitive with non-essential optimizations

Also getting wrong self.world_size when using zero3 while zero2 is alright in the same environment (8xA100) i have deepspeed==0.11.1, transformers==4.34.1, accelerate==0.24.0 and ray==2.7.1

Thanks Yes, I would be interested. Need a few more days to wrap up my current work, but I'll let you guys know when I'm ready to scope out the...

Sounds good. About to reach out this weekend. On Fri, Feb 24, 2023 at 4:50 PM Oliver Stanley ***@***.***> wrote: > Closed #1584 as > completed. > > — >...

Hello team, I’d like to work on this.