Alex
@christophschuhmann Hello! Your dataset LAION is incredible. As a researcher, I would be interested in working with your dataset in the Russian language. ruCLIP was trained on datasets from open...
@Lednik7, thank you a lot. It is advantageous! Could you create a pull request in main? I can do it by myself, but It is your contribution. appreciate it
@Lednik7 if you don't have enough time, I can do it myself - let me know
@eyaler you can use param use_cache=True with image_prompt ;) thanks a lot @oriBetelgeuse for fix here: https://github.com/sberbank-ai/ru-dalle/pull/37
@migsmurai I suppose you met this error because your GPU has not enough gRAM for VAE decoding. You can try this example, it should help: GPU 3.5Gb vRAM https://github.com/sberbank-ai/ru-dalle/pull/51 [](https://colab.research.google.com/drive/1AoolDYePUpPkRCKIu0cP9zV7lX5QGD3Z?usp=sharing) or [Jupyter version](https://github.com/sberbank-ai/ru-dalle/blob/master/jupyters/Malevich-3.5GB-vRAM-usage.ipynb): @bropines hope it helps for you
@vxzcorporation thanks for your feedback :) double BOS token will be fixed with new version of model checkpoint, sorry. if you remove it without more training, ruDALLE loses some skills...
@corranmac Thank you 🙏 I didn't test image prompt with 3.5 Gb vRAM and dwt decoder, you are right. I will create example in near future
@StrangeTcy good evening! finetuning example on free colab you can find here: **Finetuning example** [](https://colab.research.google.com/drive/1Tb7J4PvvegWOybPfUubl5O7m5I24CBg5?usp=sharing) it doesn't support DDP, but feel free add these features we would like...