-
~Parsl will hopefully supervise one intern with [Outreachy](https://outreachy.org) in the middle of 2024 on~ a project to add extra worker selection algorithms to the High Throughput Executor. This is…
-
Hi, I am measuring latency using free5GC+UERANSIM.
UERANSIM : VM1 (Virtual machine 1)
UPF : VM2
MEC : VM3
With above setup, I am measuring download speed from MEC to UERANSIM using iperf TCP.…
-
When running a model with ```--model-mode flashinfer``` (I have tested ```mistralai/Mistral-7B-Instruct-v0.2```), for a large batch (eg 50,000 text input), I usually see that the throughput is high th…
-
## Description
Hi there, I'm encountering a flickering issue when using the `Viewport::Inline` setting for the `Terminal`. On a high throughput of messages (i.e. I'm calling `terminal.insert_…
-
1. See if the benchmark setup is optimal, throughput should match standalone bench on machine. This is important since benchmark is done in a docker container.
2. Use this to more accurately detect w…
-
When scaling the batch size from 1 to a small number, say 8, I'd naively expect the generation performance to scale quite well, as we're still firmly in memory bound regime. But I observe two things …
-
### Goals
Operators should receive runtime metrics for throughput, current load, and error/success rates. These metrics can then be forwarded to a metric backend of their choice, such as Prometheus o…
-
Current implementation of sockets have a number of issues which lead to lower than possible read throughput. I have been looking at this a bit and have identified a number of improvements which are po…
-
### Nomad version
- Servers: 1.3.1
- Client: 1.3.4
### Operating system and Environment details
- 3 Servers: CentOS
- 1 Client: Windows Server 2019
- 8 cores, 24GHz
- 32GB of RAM
-…
-
There is currently one batch_size defined for both training and cross validation.
Since we do not have to keep track of gradients etc during CV, we could increase the batch_size for higher throughpu…