Jupyterlab how to

I used

module "jupyterlab" {
  count    = data.coder_workspace.me.start_count
  source   = "registry.coder.com/modules/jupyterlab/coder"
  version  = "1.0.30"
  agent_id = coder_agent.main.id
}

module "jupyterlab" {
  count    = data.coder_workspace.me.start_count
  source   = "registry.coder.com/modules/jupyterlab/coder"
  version  = "1.0.30"
  agent_id = coder_agent.main.id
}

Not sure if it create conda env, but i cant install other packages. Also i want to use my gpu in that coder space. I use kubernetes, k3s and my pods are ready

kube-amd-gpu           amd-gpu-operator-gpu-operator-charts-controller-manager-56vj26n   1/1     Running     36 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-controller-7555dfd458-jm79v                  1/1     Running     38 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-webhook-server-8549795656-dftmj              1/1     Running     23 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-gc-76ddd7ff65-5m8kj       1/1     Running     23 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-master-75649cc887-lhmb8   1/1     Running     23 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-worker-sd99w              1/1     Running     28 (26m ago)   12d

kube-amd-gpu           amd-gpu-operator-gpu-operator-charts-controller-manager-56vj26n   1/1     Running     36 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-controller-7555dfd458-jm79v                  1/1     Running     38 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-webhook-server-8549795656-dftmj              1/1     Running     23 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-gc-76ddd7ff65-5m8kj       1/1     Running     23 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-master-75649cc887-lhmb8   1/1     Running     23 (26m ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-worker-sd99w              1/1     Running     28 (26m ago)   12d

Checked docs/articles but non found. Having some for this would be great. Thank you

15 Replies

EmircanOP•6mo ago

https://github.com/ROCm/gpu-operator/issues/154

GitHub

[Feature]: RDNA 2 Support · Issue #154 · ROCm/gpu-operator

Suggestion Description I have 6800xt but it do not detect my card. Name: rocminfo Namespace: default Priority: 0 Service Account: default Node: <none> Labels: <none> Annotations: <no...

EmircanOP•6mo ago

sample pod with amd gpu

apiVersion: v1
kind: Pod
metadata:
 name: amd-smi
spec:
 containers:
 - image: docker.io/rocm/pytorch:latest
   name: amd-smi
   command: ["/bin/bash"]
   args: ["-c","amd-smi version && amd-smi monitor -ptum"]
   resources:
    limits:
      amd.com/gpu: 1
    requests:
      amd.com/gpu: 1
 restartPolicy: Never

apiVersion: v1
kind: Pod
metadata:
 name: amd-smi
spec:
 containers:
 - image: docker.io/rocm/pytorch:latest
   name: amd-smi
   command: ["/bin/bash"]
   args: ["-c","amd-smi version && amd-smi monitor -ptum"]
   resources:
    limits:
      amd.com/gpu: 1
    requests:
      amd.com/gpu: 1
 restartPolicy: Never

Phorcys•6mo ago

it doesn't seem to create a conda env, no so either you'll have to set up a virtualenv yourself or use !sudo apt install python3-<package> instead in your notebook

EmircanOP•6mo ago

okay, what about gpu passthrough? or lets say gpu scheduling in kubernetes world 🙂 I had proxmox background

EmircanOP•6mo ago

after checking scripts i see pipx does the thing

!pipx install tensorflow
!pipx inject jupyterlab tensorflow

!pipx install tensorflow
!pipx inject jupyterlab tensorflow

EmircanOP•6mo ago

          resources {
            requests = {
              "cpu"    = "250m"
              "memory" = "512Mi"
              "amd.com/gpu" = 1
            }
            limits = {
              "cpu"    = "${data.coder_parameter.cpu.value}"
              "memory" = "${data.coder_parameter.memory.value}Gi"
              "amd.com/gpu" = 1
            }
          }

          resources {
            requests = {
              "cpu"    = "250m"
              "memory" = "512Mi"
              "amd.com/gpu" = 1
            }
            limits = {
              "cpu"    = "${data.coder_parameter.cpu.value}"
              "memory" = "${data.coder_parameter.memory.value}Gi"
              "amd.com/gpu" = 1
            }
          }

not sure but could be driver issue..

EmircanOP•6mo ago

matifali•6mo ago

Usually the expectation is that the environment(workspace) is pre-configured with GPU access and any other tools i.e. pytorch etc. The module only starts (and optionally installs) the jupyterlab ik the workspace

Phorcys•6mo ago

(@emircanerkul )

EmircanOP•6mo ago

couldn't get it @Phorcys I already installed the module also spesified gpu labels as like i did in other outside coder pods but getting that error. second run i only see cpus not gpu

Phorcys•6mo ago

would you be able to share your base image?

EmircanOP•6mo ago

Sure; here all tf code.

message.txt

EmircanOP•6mo ago

codercom/enterprise-base:ubuntu

Phorcys•6mo ago

hey, sorry for the late reply have you installed the AMD GPU drivers on your Kubernetes node(s)? if not, please install them and check via rocm-smi if yes, then i believe the container also needs to be compatible, so maybe try using the rocm/dev-ubuntu-22.04:latest base image instead

EmircanOP•6mo ago

No worries and thank you, its just side research and not an urgent thing. I already install all amd things and i thought labeling make all work automatically, Kube node feature discovery should do the thing automatically i thought but yea might not enough, yea probably need to merge codercom/enterprise-base:ubuntu with rocm/dev-ubuntu-22.04:latest because replacing it didnt worked

kube-amd-gpu           amd-gpu-operator-gpu-operator-charts-controller-manager-56ldnkx   1/1     Running             11 (2m47s ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-controller-7555dfd458-4fsql                  1/1     Running             11 (2m47s ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-webhook-server-8549795656-4t5p5              1/1     Running             10 (2m47s ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-gc-76ddd7ff65-5m8kj       1/1     Running             33 (2m47s ago)   25d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-master-75649cc887-lhmb8   1/1     Running             33 (2m47s ago)   25d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-worker-sd99w              1/1     Running             39 (2m47s ago)   25d
kube-amd-gpu           default-deviceconfig-device-plugin-xgwqv                          1/1     Running             9 (2m47s ago)    12d
kube-amd-gpu           default-deviceconfig-metrics-exporter-sgz8d                       1/1     Running             68 (2m47s ago)   12d
kube-amd-gpu           default-deviceconfig-node-labeller-bks4g                          1/1     Running             9 (2m47s ago)    12d
kube-amd-gpu           default-deviceconfig-test-runner-qbfkn

kube-amd-gpu           amd-gpu-operator-gpu-operator-charts-controller-manager-56ldnkx   1/1     Running             11 (2m47s ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-controller-7555dfd458-4fsql                  1/1     Running             11 (2m47s ago)   12d
kube-amd-gpu           amd-gpu-operator-kmm-webhook-server-8549795656-4t5p5              1/1     Running             10 (2m47s ago)   12d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-gc-76ddd7ff65-5m8kj       1/1     Running             33 (2m47s ago)   25d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-master-75649cc887-lhmb8   1/1     Running             33 (2m47s ago)   25d
kube-amd-gpu           amd-gpu-operator-node-feature-discovery-worker-sd99w              1/1     Running             39 (2m47s ago)   25d
kube-amd-gpu           default-deviceconfig-device-plugin-xgwqv                          1/1     Running             9 (2m47s ago)    12d
kube-amd-gpu           default-deviceconfig-metrics-exporter-sgz8d                       1/1     Running             68 (2m47s ago)   12d
kube-amd-gpu           default-deviceconfig-node-labeller-bks4g                          1/1     Running             9 (2m47s ago)    12d
kube-amd-gpu           default-deviceconfig-test-runner-qbfkn

Learning a lot more today. - Enabled registery: https://docs.k3s.io/installation/registry-mirror - Forked your coder docker repo - Based on FROM rocm/dev-ubuntu-24.04:latest - Build via docker, saved image and imported via k3s ctr images import https://www.geekandi.com/2023/02/17/import-docker-image-into-k3s/ Tested all looks good but jupyterlab tensorflow still give same error. I'll try my luck with https://hub.docker.com/r/rocm/tensorflow/tags k exec coder-dd716dea-c5a1-4351-87c6-9d4c1efa7aa2-797b7674db-z4xnh -n coder -- sudo rocminfo

*******                  
Agent 2                  
*******                  
  Name:                    gfx1030                            
  Uuid:                    GPU-df8c88475109d004               
  Marketing Name:          AMD Radeon RX 6800 XT  
...

*******                  
Agent 2                  
*******                  
  Name:                    gfx1030                            
  Uuid:                    GPU-df8c88475109d004               
  Marketing Name:          AMD Radeon RX 6800 XT  
...

I build with tensorflow 22gb image size 🤯 but still getting same issue. I think it might be related with user group access. https://discord.com/channels/747933592273027093/1370758361108447272 Of course .. not things always goes good https://github.com/pypa/pipx/issues/1635 Here is pr https://github.com/coder/images/pull/296 (its quite niche area and could be waste of build credits but incase you wanna add *but i want to test it more)

Gaming

Programming

Jupyterlab how to

Did you find this page helpful?