Open Bortus-AI opened 6 months ago
I'd like to hide this
and just show this one without having to arrow over
Hi, this should definitely be a feature - can you please send me the output of:
docker exec CONTAINER yarn cli raw-data --gpu
If you are not using docker, just omit the
docker exec CONTAINER
part.
yarn cli raw-data --gpu
I also would love to see this added! My Proxmox VM's virtual gpu (?) is displayed before the passed through P4000
const gpuInfo = {
controllers: [
{
vendor: '',
subVendor: '',
model: '',
bus: '',
busAddress: '00:02.0',
vram: 4,
vramDynamic: false,
pciID: ''
},
{
vendor: 'NVIDIA Corporation',
model: 'GP104GL [Quadro P4000]',
bus: 'Onboard',
busAddress: '00:10.0',
vram: 8192,
vramDynamic: false,
subVendor: 'Hewlett-Packard Company',
driverVersion: '545.29.02',
subDeviceId: '0x11A3103C',
name: 'Quadro P4000',
pciBus: '00000000:00:10.0',
fanSpeed: 50,
memoryTotal: 8192,
memoryUsed: 2,
memoryFree: 8111,
temperatureGpu: 43,
powerDraw: 6.91,
powerLimit: 105,
clockCore: 139,
clockMemory: 405
}
],
displays: []
}
My compose file:
version: '3'
services:
dashdot:
image: ghcr.io/mauricenino/dashdot:nvidia-5.8.3
ports:
- $PORT:3001
volumes:
- /:/mnt/host:ro
privileged: true
deploy:
resources:
reservations:
devices:
- capabilities:
- gpu
environment:
- DASHDOT_SHOW_HOST=$DASHDOT_SHOW_HOST #true/false
- DASHDOT_ENABLE_CPU_TEMPS=$DASHDOT_ENABLE_CPU_TEMPS #true/false
- DASHDOT_ALWAYS_SHOW_PERCENTAGES=$DASHDOT_ALWAYS_SHOW_PERCENTAGES #true/false
- DASHDOT_ACCEPT_OOKLA_EULA=$DASHDOT_ACCEPT_OOKLA_EULA #true/false
- DASHDOT_CUSTOM_HOST=$DASHDOT_CUSTOM_HOST #String
- DASHDOT_WIDGET_LIST=$DASHDOT_WIDGET_LIST #refer to https://getdashdot.com/docs/config/ui-features#dashdot_widget_list
Hello, dear MauriceNino,
I have a similar need. My Dashdot is running inside a Docker container with an NVIDIA environment. Dashdot is working very well and has detected my Tesla P4 and its associated data. However, for some reason, when using the API call /load/gpu, the GPU list is: {"layout":[{"load":0,"memory":0},{"load":95,"memory":61}]}, which places an empty GPU before my Tesla P4. A small feature to switch the GPU reporting order or the filter functionality you mentioned in issue #290 be sufficient for me.
Below is the output of running yarn cli raw-data --gpu inside the Docker container. I would be very grateful for any assistance you can provide.
const gpuInfo = { controllers: [ { vendor: '', subVendor: '', model: '', bus: '', busAddress: '00:01.0', vram: 4, vramDynamic: false, pciID: '' }, { vendor: 'NVIDIA Corporation', model: 'GP104GL [Tesla P4]', bus: 'Onboard', busAddress: '01:00.0', vram: 8192, vramDynamic: false, subVendor: 'NVIDIA Corporation', driverVersion: '535.161.08', subDeviceId: '0x120710DE', name: 'GRID P4-8Q', pciBus: '00000000:01:00.0', memoryTotal: 8192, memoryUsed: 1210, memoryFree: 6325, clockCore: 885, clockMemory: 2999 } ], displays: [] }
Description of the feature
I would like to hide a specific GPU and only show my Nvidia P4 card and not the others. Correct me if I am mistaken and this is already an option that I missed. It would also be nice to be able to set the order of which card appears first in the UI.
Additional context
No response