-
I tested various broadcasting settings against all pointwise binary ops, and it produced interesting results. I tested both uni- and bidirectional broadcasting by using the following input shapes:
`…
-
It seems that pytorch breaks down `aten.upsample_nearest2d` and `aten.upsample_bilinear2d` into multiple more generic ops first before passing to ttnn backend. Therefore in order to get good performan…
-
### Issue
Dispatch time can be a limiting factor for perf, especially for decode, where op device latency are low. We see that ops generally have 3k to 6k cycles of dispatch time, regardless of the d…
-
![图片4](https://github.com/user-attachments/assets/656120c2-d5de-4672-920f-5e9e827f9f56)
[WRN] Health check update: 10 slow ops, oldest one blocked for 56 sec, mon.d has slow ops (SLOW_OPS)
9/29/…
-
C++17 needs additional comparison operator overloads with reversed parameters. For C++20 and newer it is unnecessary due to "synthesized three-way comparison" feature.
-
Canonical ops: https://docs-preview.pytorch.org/90644/ir.html
The list below is an incomplete list of canonical ops + other ops ordered by usage. canonical ops are marked with `canonical`. Be sure…
-
Occurring in self-attention, mostly around stack/slice op decompositions:
-
Fold/Unforld it torch and their analogs space_to_depth/depth_to_space in TF sometimes used in models (e.g. https://paperswithcode.com/paper/selfreformer-self-refined-network-with )
It would be grea…
-
### Descrição da tarefa
Fazer um estudo sobre os botões desabilitados do Mística
---
### Data de abertura
15/04/2024
### Data de início
### Data de entrega
-
So far, we have assumed that `--benchmark-ledge-ops` reports relevant elapsed times. Issue https://github.com/IntersectMBO/ouroboros-consensus/issues/223 for example identifies that as a risk.
Whil…