-
Hi, I'm benchmarking flashinfer on H100, and I'm running attention for the decoding stage.
I use q_head = kv_head = 40, which is the standard attention for llama 13B.
I tried use_tensor_cores = …
-
Upon upgrading our Azure CI machines to use the new Azure Cobalt ARM64 processors, we started seeing frequent compiler crashes when building a large Swift project. After some investigation, the culpri…
-
Everything runs perfectly except one thing: pipe's output stream doesn't get closed when pipe's input stream is closed. And that causes the program to run forever until it's closed manually by Ctrl+C.…
-
https://github.com/user-attachments/assets/4d9c680f-53bd-49ac-a499-ff4bb382ecb7
Press Super + X, right window becomes small pop up.
swap left window to right.
Try to move pop up window …
-
## Description:
Extension is unable to activate unnamed conda environments created using the `conda create -p` command **after the latest update**. Previously, these environments were activated cor…
-
### Checks
- [X] I have checked that this issue has not already been reported.
- [X] I have confirmed this bug exists on the [latest version](https://pypi.org/project/polars/) of Polars.
### Re…
-
data = snap.read_dataset("combined.h5ads")
data
AnnDataSet object with n_obs x n_vars = 277377 x 526765 backed at 'combined.h5ads'
obs: 'sample', 'leiden'
var: 'count', 'selected'
uns…
-
So, I have this Pisces P100 miner that has been working completely ok for the better part of last year. A month ago miner's dashboard started showing blank witness logs, miner height is also blank, it…
-
~~(without small chunk workaround)~~ we don't need reverting that workaround to reproduce the failure
Intermittently
-
### Problem description
Write only one dictionary when sinking to IPC.
It would be great that when writing categorial data to an IPC sink, only one unified dictionary is written, instead of multip…
ghuls updated
10 months ago