[bug]: "Retry install" in model manager fails for gated models requiring a Hugging Face token after adding it #9829

Is there an existing issue for this problem?

• I have searched the existing issues

Install method

Invoke's Launcher

Operating system

Linux

GPU vendor

Nvidia (CUDA)

GPU model

RTX 4090

GPU VRAM

24GB

Version number

Version 7.0.0-alpha.2

Browser

Vivaldi 8.2.4133.84 / Chromium 152.0.7977.160

System Information

No response

What happened

When trying to download a gated model requiring an access token via Invoke's model manager (e.g. starter models), it naturally fails when no token is provided under the API keys. However, if a valid token is added after a failed attempt and the button "Retry install" in the install queue is pressed, the attempt still fails with the same error, saying that no token is available. Only when starting a completely new download attempt from the model manager (not the queue) does it properly pick up the access token.

What you expected to happen

That the "Retry install" function recognizes the changed / added token and uses it accordingly for the retry attempt.

How to reproduce the problem

1.) Try downloading a gated model without a Hugging Face token - e.g. FLUX.1-schnell_ae for a quick test. For testing, it also works by removing any existing Hugging Face from the API keys before the download attempt.

2.) Wait for the error saying that the model is gated and needs an access token.

3.) Add a valid Hugging Face token under the API keys, expand the install queue and click the "Retry install" button. The download will still fail.

Additional context

No response

Discord username

No response