Finetuned Model Deployment Help

[Aayushma_Sharma]

Hello there,

I am trying to see the performance of the model that I finetuned recently. I trained SmolVLM2 LoRA. I wanted to see how well it performs. Is there any way to try it? I tried deploying through the workspace; however, I cannot find my model there. I wonder why is that so?

Could you please help me out. Looking forward to your response. Thank you.

• Project Type: Multimodel

• Operating System & Browser: Windows, Microsoft Edge

• Project Universe Link or Workspace/Project ID: trial_2-text-image-pairs-ruxwy/

  

[Ford]

Hi @Aayushma_Sharma!
The best way to deploy a finetuned SmolVLM2 for a production environment is via a dedicated deployment. Here is the Roboflow guide to Dedicated Deployments.

Additionally, here is a great blog post on fine tuning and deploying a SmolVLM. How to Fine-Tune a SmolVLM2 Model on a Custom Dataset

[Aayushma_Sharma]

Hey, thank you for your response. I followed the same blog post (I followed the YouTube video posted by Roboflow).

However, the main problem I am having is that I cannot find the model I trained in the SVOLM block in my workspace.

I wonder what’s causing this problem. I have already trained the model twice, I cannot see any of the trained model. I would appreciate if you could provide any help on this matter. Thank you!

[Aayushma_Sharma]

@Ford Could you please provide some suggestions?

[Aayushma_Sharma]

If I added the model ID of my finetuned model. I received server error. Do you have any suggestions to solve this error? Or any alternative ways to test and/or deploy our model?

[Aayushma_Sharma]

Using the deployment code snippet provided by Roboflow. I again got the same error.

[Automatez]

I can at least answer that for the Workflow block diagram you showed - it looks like you are pulling in the pre-built “SmolVLM2” block that uses a fully trained model by default. Pretty sure there would not be a way in that block to replace it with your fine-tuned version.

[Aayushma_Sharma]

Thank you for your reply! @Automatez
Do you have any suggestions on how to test the fine-tuned SmolVLM model?

[Automatez]

Try to start a new workflow and again click on “Add a Model”.

But now try selecting the “Your Models” tab and finding your fine-tuned model. (I think maybe you selected “Model Type > SmolVLM2” last time.)

[Aayushma_Sharma]

It worked! Thank you very much. I appreciate your help! @Automatez

Yeah, I was using the SmolVLM2 so I could not use the model I trained.