Make a prediction and plot the results after training

[brunodifranco]

I've finished training the model (with GPT2), and I have the model saved, as shown below:

How can I load to model back, make a prediction and then plot ? like so:

[59592-9]

Hello, can their model be successfully configured? If possible, what is the minimum level of GPU required

[brunodifranco]

Hello, can their model be successfully configured? If possible, what is the minimum level of GPU required

I managed to run it in a RTX 3050 Laptop GPU 4GB, but I had to tweak a bunch of parameters and use GPT2, instead of bigger models.

[Huang-Jingxiang]

Maybe this will help. https://towardsdatascience.com/time-llm-reprogram-an-llm-for-time-series-forecasting-e2558087b8ac

[JackKoLing]

Keep an eye on the demo version

[lehazare]

Hello @brunodifranco, do you think it would be possible for you to share the GPT2 tweak you mentioned ? @JackKoLing what do you mean ?

[JackKoLing]

Hello @brunodifranco, do you think it would be possible for you to share the GPT2 tweak you mentioned ? @JackKoLing what do you mean ?

I mean waiting for the plot demo..

[JackKoLing]

I've finished training the model (with GPT2), and I have the model saved, as shown below:

How can I load to model back, make a prediction and then plot ? like so:

just get all input batch, GT, and outputs on test datasets, and then you can get the MAE, MSE and plot by batch_x, GT, and outputs.