Thank you for sharing @lucidrains ! Really fantastic work. I watched Yanniks video on perceiver and this repo really helps to go one or two levels deeper and understand the paper.
However, for applying this new architecture to practical applications, most of us without 100s of TPUs at our disposal, require pretrained models on imagenet, audioset etc. Could anyone that has successfully trained the model, can share weights?
Thank you for sharing @lucidrains ! Really fantastic work. I watched Yanniks video on perceiver and this repo really helps to go one or two levels deeper and understand the paper. However, for applying this new architecture to practical applications, most of us without 100s of TPUs at our disposal, require pretrained models on imagenet, audioset etc. Could anyone that has successfully trained the model, can share weights?