Mathias Gatti

I'm trying it in Google Colab, It doesn't seem to be loading the model successfully, anyone experiencing a similar problem? After training the model a little bit ```python from mingpt.trainer...

Amazing! Thanks, I will test it, anyway checking it quickly I don't see a difference between your `save_checkpoint` implementation and the one from [this repository](https://github.com/karpathy/minGPT/blob/master/mingpt/trainer.py#L57). Am I missing something? By...

oh okay, I saw you used a deepcopy while loading the checkpoint, maybe that helps, I will try that.

:'( It's still failing to load the model successfully, I created [this](https://colab.research.google.com/drive/1sjdVtK6fD7TnG_WLYO-xbGwXB8RpDg82?usp=sharing) simple colab to replicate my problem.

It's weird, I don't see any difference with this [colab](https://colab.research.google.com/drive/1sjdVtK6fD7TnG_WLYO-xbGwXB8RpDg82?usp=sharing#scrollTo=c_11M1ygHmoh), it saves the model while training like [this](https://github.com/karpathy/minGPT/blob/master/mingpt/trainer.py#L57). And it loads the model as you suggested.

Thank you very much! I tried your colab with my shakespeare txt and it loaded the model successfully :D By the way have you uploaded the midi composition results somewhere?...

[Here](https://colab.research.google.com/drive/1zl787bEG-aMTxUU5reDxfl8hR62EIdji?usp=sharing) is the colab modified a little bit for regular text generation

utf8 already supports english, are you experiencing issues if you just try to open the file with utf8?

Not sure, the colab I shared used to allow training, saving the checkpoint and loading it again. I didn't try training again from that

Hey thank you very much for the answer. Did you try those solutions on colab? I'm trying it but didn't work. About the 100 GB of video memory I thought...