Shiyan Deng

[Enhancement] Allow Bert Encoder to specify hidden dime for the fc layers

6

This was the second enhancement in https://github.com/NVIDIA/FasterTransformer/issues/98. Add support for different hidden dimension size in bert's fc layers. Currently in bert encoder feed forward part, the hidden dim of the...

use gpu_library_selector for permute_pooled_embedding_ops

1

Summary: X-link: https://github.com/pytorch/FBGEMM/pull/2310 ATT Differential Revision: D53364919

hipify torchrec/inference

1

Differential Revision: D53434630

add autodep rule for torchrec/inference/Batching.h

2

Summary: Allow autodeps to auto add `fbcode//torchrec/inference:batching` to the buck target when we imported `torchrec/inference/Batching.h` Reviewed By: houseroad Differential Revision: D54284062

add a new function to update sparse delta

2

Summary: The only difference is that in the new function, we don't pass in a weight placement tensor. Instead we only pass in a weight placement variable since weight placement...

[Feature]: ECC Mode Check

1

### Suggestion Description Hi, for nvidia gpu we can use nvidia-smi to check ECC mode, and use the tool to turn ECC on/off. I couldn't find any documentation on AMD...

## Purpose We have draft model trained in different format compare to the target model and need to use a different loader to load it. Adding the support for this....