Ilya Zaychuk

Thank you for you work! Is it possible to train HybrIK-X model now?

about training (video/photo) and different size of betas in datasets

10

1. I saw on the main page, that SMPLer-X inference scripts expect video data as input. Is it possible to modify model to support single images? 2. Do you train...

Very big MAE SMPL-X Translation between GT and Predicted

1

Thank you for your research. I trained model with custom dataset, it has low PA-MPVPE and PA-MPJPE errors, but absolute mean translation error is about [0.5, 0.5, 40] XYZ... Why...

Training Code

3

Thank you for your research. Do you have any plans to release training and data processing code?

alibi_slopes with flash_attn_varlen_kvpacked_func

6

Hi, is it possible to make an `alibi_slopes` attention bias matrix with `flash_attn.flash_attn_varlen_kvpacked_func` look like this for one batch?  According to the documentation, the `alibi_slopes` tensor should have a...

Hi, thank you for your great work. Could you share your successful train logs any of your model? I would like compare with my train process.

Hi, thank you for you work. I found incorrect annotations in these sequences: - **p000451_a001082** - **p000455_a001244** - **p000456_a000064** - **p000519_a000226** - **p000520_a000226** - **p000533_a000197** - **p000539_a000279** - **p000545_a005731** -...

Hi, Thank you for you paper. Could you share CMU Panoptic annotations?

Send same messages to WebSocket server

2

Hello, I need to send status messages via **WebSocket** (for example: `"client_is_ready"`), which tell server that's dash-app is ready to receive new message. And _**dash-extensions**_ doesn't allow this. Here is...

SparseMaxPool3d returns Zero if all sparse elements are negative

1

spconv.SparseMaxPool3d returns zero if all elements in the kernel window are negative (kernel is not full), but should return the minimum negative value.