Lesson 2 official topic

I had some similar problems before(not exactly the same though) with Paperspace.
I end up having to run the file in jupterlab on the left of the panel, and my problem is miraculously solved…

2 Likes

Supporting JoshL’s answer, the JupyterLab button is boxed in red…
image

2 Likes

Thanks! It’s worked. I’m on my way.

Thanks!

Edit: solved!
I was not inspecting correctly (its been a while).
solution to display confidence as well:

const json = await response.json();
    console.log("json", json);
    const label = json['data'][0]['confidences'][0]['label'];
    const conf = json['data'][0]['confidences'][0]['confidence'];
    // console.log("conf", conf);
    results.innerHTML = `<br/><img src="${reader.result}" width="300"> <p>${label}</p> <p>${conf}<p>`

Original question:
Hello, I successfully cloned the tiny pets website. My goal is to add the confidence level to the single, multiple, and parallel page results. Starting with the single inference page, I am struggling to add another line that displays the confidence of the inference.

Among other things, the closest I have gotten is by adding a paragraph div such as:

    const json = await response.json();
    // console.log("json", json);
    const label = json['data'][0]['confidences'][0]['label'];
    const conf = json['data'][0]['confidences'][0];
    console.log("conf", conf);
    results.innerHTML = `<br/><img src="${reader.result}" width="300"> <p>${label}</p> <p>${conf}<p>`

Result:

The console log does not help as expected. I am not sure how to dig into it deeper to get the confidence to display. Any advice would be greatly appreciated!!

Edit: solved!

Hello beautiful Ai community! :smile:

First, I want to say that I’m loving the course so far … In just a lesson an a half, I managed to put together a model that is able to distinguish between my two children, and I’m in awe of it all :+1:

I do have a question related to using learn.export(). I get an error when I try to do this:
RuntimeError: don't know how to determine data location of torch.storage._UntypedStorage

My environment is running on an Intel MacBook Pro 16" 2019 (macOS Monterey) in Visual Studio Code.
The Python version I’m using is v3.10.6
PyTorch version is v1.12.1
and everything else is working fine, but for some reason I cannot run learn.export() :thinking:

If anybody has any insight as to why this is, it would be much appreciated.

Thank you in advance.

Mac support in PyTorch is still very early and in fastai even earlier still – I suspect you might have just found something that’s not working in PyTorch yet.

Thanks :pray: for the reply :+1:

Thanks a lot, Dylan

Hello everyone,
I have been making a model to predict two types of wasps that exist and coexist in my locality. The first thing I did was to download the images with the search_image_ddg function. Next, I prepared the images in a datablock. When I was about to train the model I realized that one of the terms I was searching for was wrong. I retraced my steps to re-download all the images again. But the datablock now only handled the data downloaded the first time, i.e. images that did not correspond to the term I was searching for the second time.
My question is, how can I delete the downloaded and restart the process?
Thanks

Hi Silvino,

What are you using to run this Notebook, Jupyter Notebook, or some other kind of app?
From what I can tell, normally you should see a directory structure appear in the Home Page tab of Jupyter Notebook and you could navigate to the location of the Notebook itself, there you should see the directory structure that contains the images downloaded. :face_with_monocle:

All you have to do then, is just go through them and get rid of the ones you don’t want.

Ultimately you could just delete them all together and re-run the Notebook’s cells to re-download them.

Hope this helps :crossed_fingers:

I’m using Colab. But i cannot delete the folders from the own colab becouse this folders have to be empty and my hard disk, I can’t find them

I understand, I know that when using Google Colab all of the downloaded files reside on their server, so I get the confusion.

Have a look at:

Hopefully it will provide some help :slightly_smiling_face:

But I suspect that you could just open another Google Colab space (tab) and re-use the code to run a fresh instance :face_with_monocle:

1 Like

Hi Fast.ai community,
I tried to upload the export.pkl file with load_learner(export.pkl) but it return an error:

AttributeError: Custom classes or functions exported with your Learnerare not available in the namespace currently. Please re-declare or import them before callingload_learner`:

full msg:

any ideas how I could fix this problem?
thank you all.

It would help to know what platform are you on!

Generic advice… Jump into the command shell and delete from there,
or from a notebook use “!rm -r {yourdownloadpath}”


Why did i get this error ? anyone could help?


I tried to implement my own application but it has been building forever. I copied and pasted the code exactly and verified it as well. Did anybody else face this issue?

You’re probably missing the requirements.txt file.

Hello. I am having this problem. Can anyone help?

Hi jeremy…yes indeed, I dont’ have that. I typed the code instead of exporting like you showed because I had trouble exporting the code. So this txt file comes with your export setting or it’s Huggingface that failed to generate ?
Thanks