-
**Is your feature request related to a problem? Please describe.**
It is very difficult to build the oss-module.so and integrate it with the redis-stack-server. On top of it this bears the questions:…
-
Dapr middleware is not part of the list of supported dapr components but dapr middleware component definition is accepted.
I am particularly interested in the ratelimit and bearer-token middleware.
…
-
Currently we have worker-level resources, which are useful for limiting access to things that are limited per-worker (memory, GPU, etc...). Sometimes tasks will consume resources that are globally lim…
-
When you do:
`PrincipalsDAO.getPrincipals(principalIds, ['principalId'], function() { ... });`
The result will be different based on if there was 1 or multiple `principalIds` in the array of the fir…
-
The rate limit plugin requires us to follow one of the options:
**First.** To store a distributed CRDT counter which is merged via clustering protocol on an event or timing basis.
Pros:
- Each …
-
### 🐛 Describe the bug
I was trying to use CUDA graphs (`torch.cuda.make_graphed_callables`) on a model wrapped with `FullyShardedDataParallel` (FSDP) and I got the following error:
```
-- Process …
-
In certain applications, for example in [single particle analysis](https://en.wikipedia.org/wiki/Single_particle_analysis), the geometry is well-known but does not fit into any of the currently suppor…
-
## Problem
The starter is currently relying on REST API of k6 to start the runners. It is a workable solution that allows to synchronize the start of the test. But it requires to schedule a separat…
-
### 🐛 Describe the bug
GPU: 8*A6000
CUDA Version: 11.7
Python Version: 3.8.10
colossalai Version: 0.2.8
when I train PPO by
```
torchrun --standalone --nproc_per_node=8 train_prompts.py \
…
-
# Load Balancing Algorithms
Different algorithms can be used to distribute traffic among backend servers:
1. **Round Robin:** Cycles through a list of servers and forwards each new request to the ne…