-
A clear and concise description of what the bug is.
After running Keda stably for over 2 years with the cert-manager integration we found a subtle but important bug that essentially turns into a ex…
-
### Describe the bug
After creating a node group using a custom AMI (new feature as of 2024.9.1), I have not been able to successfully launch a JupyterLab Pod on the new image
Note that the ami_…
-
**Describe the bug**
The Checkbox component significantly slows performance & rendering time at scale (2000+ checkboxes) and slows functionality, like checking or unchecking a box. A difference of 4.…
-
### What happened + What you expected to happen
We have an object detector deployment that we restart every few minutes to prevent memory leaks in third party code from adding up. The way we restart …
-
### What's the problem?
Equinox v0.11.8 causes various TypeErrors in the score matching code. See https://github.com/gchq/coreax/actions/runs/11437260050/job/31816415291?pr=816, or the traceback from…
-
https://github.com/redis/redis/releases/tag/8.0-m01
New commands:
>- 7 new data structures: JSON, Time series, Bloom filter, Cuckoo filter, Count-min sketch, Top-k, t-digest
>- Redis scalable que…
-
scaling works but it's not very user friendly, perhaps some styling could fix this
Tammo: I looked at it, tried and failed finding a solution
Suggestion: integrate in the image dialogue a width fi…
-
我下载了WuKong数据集，有效图片约700万条，然后基于此训练vit-base-16这个模型，使用常规transformer，按照论文stage1-stage2步骤训练，都是正常的。但是使用flash-attention却出现loss不下降，acc也不增加问题。
- 使用常规transformer的训练日志如下：
```
024-03-08,09:41:15 | INFO | Rank…
-
**Which component are you using?**:
cluster-autoscaler
**Is your feature request designed to solve a problem? If so describe the problem this feature should solve.**:
The problem …
-
...with the caveat that label scaling is not veridical to utterance duration
E.g. this:
```this_conv |>
ggplot(aes(x = line_end, y = line_participant)) +
ggtitle("Relative frequency of ele…