-
# Motivation
On Hopper, [efficient gemm requires warp-specialization](https://github.com/NVIDIA/cutlass/blob/main/media/docs/efficient_gemm.md#warp-specialization), which is not currently supported…
-
### Apache Iceberg version
1.5.0
### Query engine
Flink (iceberg-flink-1.18:1.5.0)
### Question
Hello, I'm using iceberg-flink-1.18-1.5.0.
I've configured the [flink-operator autoscaler fe…
-
`--num-gpus` is implemented by [sharding each expert layer across GPUs](https://github.com/open-compass/MixtralKit/blob/38bbb5524ee6dcecd2d4724b06179c6783019db4/mixtralkit/layers/moe.py#L27), i.e. exp…
-
As a Terraform User, I should be able to specify a `parallelism` count on resource instances containing a `count` parameter So That I Can handle creating API resources via providers at a more sane rat…
-
**Describe the situation**
We have a query with many `ALL LEFT JOIN`:
```
SELECT * FROM
t1
ALL LEFT JOIN t2 ON ...
ALL LEFT JOIN t3 ON ...
ALL LEFT JOIN t4 ON ...
```
t1 is a memory table…
-
Hi, I have a code in which the most computationally expensive part is the contraction of the whole network, that is composed by a number of tensors usually larger than 15/20. I am using the functions …
-
## What problem are you trying to solve?
The vast majority of Maven builds I've encountered are single-threaded - quite a waste of processing power in today's multi-core world. While not every mul…
-
Dear @phoboslab ,
Thanks for this great project.
To make the decoder more performant, there are a few areas to explore.
1. Enable the web workers to offload the decodePicture works to another thr…
-
Right now is table to avoid redundant reads.
By splitting data on memory we could provide more fine-granular parallelism, i.e. per column, while still avoiding redundant reads.
-