fix: upload model on podman machine on WSL to speed loading

[lstocchi]

So, it should be ready.
When you are on WSL the model gets uploaded to the podman machine. This makes us loses like 20sec during the start of application but it makes us gain minutes when the model is actually loaded by the model service.

So to test it,

1. run an application
2. when it reaches the "application is started" state, if you see the model service logs, it should be already running with the model loaded
3. if you check inside the podman machine you should see the model
   podman machine ssh ls

If the user system is no Windows, it works as before. Qemu will be added on another PR

[jeffmaury]

The only concern I have is that it uses the default Podman machine which may be stopped and deployments are performed against the first Podman machine that is running

[lstocchi]

@jeffmaury updated

[axel7083]

I don't understand why you renamed from ProgressEvent to ProgressiveEvent ?

[lstocchi]

they are two different things.
Ai studio had the DownloadEvent. This PR is adding an upload action which would require a similar event and so i renamed the DownloadEvent to ProgressiveEvent. It's like downloadEvent and uploadEvent = progressiveEvent

[axel7083]

I feel like Progress and Progressive are too close and could lead to some confusion, we might want to find something different to improve readability.

BaseEvent instead of DownloadEvent and keep ProgressEvent to talk about the real progress.

or

TransferEvent if we want to keep some action/wording of what is happening

I have a preference for BaseEvent, as we could create its own IBaseEvent class, and keep it separate from the logic of the Downloader/Uploader

[lstocchi]

progressing = happening or developing gradually or in stages. Look to me that it expresses well what it does and it can be reused.
By using TrasferEvent we are binding it only to transferring actions which may not be true in future, no?
Maybe ExecutingProgressiveEvent instead of ProgressProgressiveEvent?

[axel7083]

here is the Download event

export interface DownloadEvent {
  id: string;
  status: 'error' | 'completed' | 'progress' | 'cancelled';
  message?: string;
}

This seems very basic and generic, I still think something short and easy to read like BaseEvent would be more suitable than ExecutingProgressiveEvent... but this is just my opinion, if other people are fine with your choice I open to accept it !

[lstocchi]

@jeffmaury rebased

[axel7083]

While starting an Application I got the following

AI Studio	Pod details

Is this expected ?

[lstocchi]

Is this expected ?

No, have you done something in particular? I'll try now

[lstocchi]

Did you push your commit ?

🤦 no, sorry. Pushed now

[axel7083]

Did you push your commit ?

🤦 no, sorry. Pushed now

lucas-pr.mp4

Maybe my repo is too old ? (will retry after I ate)

[lstocchi]

Maybe my repo is too old ? (will retry after I ate)

You can try, the error seems related to the podman machine though 🤔
What if you run podman machine list --format json? Does it work fine? Do you have any value in the podman cli path setting?

I'll add a check to prevent this kind of error

[axel7083]

Maybe my repo is too old ? (will retry after I ate)

You can try, the error seems related to the podman machine though 🤔 What if you run podman machine list --format json? Does it work fine? Do you have any value in the podman cli path setting?

I'll add a check to prevent this kind of error

PS C:\Users\axels> podman machine list --format json
[
    {
        "Name": "podman-machine-default",
        "Default": false,
        "Created": "2023-09-23T19:18:48.434351+02:00",
        "Running": true,
        "Starting": false,
        "LastUp": "2024-03-12T12:55:57.0740947+01:00",
        "Stream": "35",
        "VMType": "wsl",
        "CPUs": 12,
        "Memory": "689012736",
        "DiskSize": "72994521088",
        "Port": 64772,
        "RemoteUsername": "user",
        "IdentityPath": "C:\\Users\\axels\\.ssh\\podman-machine-default",
        "UserModeNetworking": false
    }
]

[lstocchi]

Maybe my repo is too old ? (will retry after I ate)

You can try, the error seems related to the podman machine though 🤔 What if you run podman machine list --format json? Does it work fine? Do you have any value in the podman cli path setting?
I'll add a check to prevent this kind of error

PS C:\Users\axels> podman machine list --format json
[
    {
        "Name": "podman-machine-default",
        "Default": false,
        "Created": "2023-09-23T19:18:48.434351+02:00",
        "Running": true,
        "Starting": false,
        "LastUp": "2024-03-12T12:55:57.0740947+01:00",
        "Stream": "35",
        "VMType": "wsl",
        "CPUs": 12,
        "Memory": "689012736",
        "DiskSize": "72994521088",
        "Port": 64772,
        "RemoteUsername": "user",
        "IdentityPath": "C:\\Users\\axels\\.ssh\\podman-machine-default",
        "UserModeNetworking": false
    }
]

Only that machine? I wonder why you don't have it set to default. Podman desktop should ask you to set the active machine as the default one.

[axel7083]

Only that machine? I wonder why you don't have it set to default. Podman desktop should ask you to set the active machine as the default one.

Apparently it is not, but this case should be handle I guess

[lstocchi]

Apparently it is not, but this case should be handle I guess

I was able to replicate your case by using rootful machine.
So now i use the running connection name to recreate the podman machine, it should work.
Whenever you have time, please give it another try. Thanks

[axel7083]

In another PR, that would be better to have this as a callable method, so we could allow from the model list to upload it to the podman machine manually by the user. We would also probably want to be able to delete them

[lstocchi]

yes, I agree 👍

[axel7083]

Apparently it is not, but this case should be handle I guess

I was able to replicate your case by using rootful machine. So now i use the running connection name to recreate the podman machine, it should work. Whenever you have time, please give it another try. Thanks

Once uploaded, the model is soo fast to start !! Love it.

[jeffmaury]

LGTM. Great improvment

[axel7083]

Just tested and works fine ! Great job 👏 !!