Romain Beaumont

https://colab.research.google.com/drive/10P81dVS7YKCMUHF3FA7WD3Q_mp-cCWIA#scrollTo=VVElbFFcb5T7 new eval with the new checkpoint https://huggingface.co/krish240574/Dalle2-Diffusion-Prior/blob/main/1651473037.600823_saved_model.pth now it works, we improve the similarity! previous checkpoint was 0.27 -> 0.09 now 0.27 -> 0.28 we're going to make a...

this is almost done, zero shot eval might be the last thing here, some people are on it

ok the biggest I can fit is Number of parameters: 1639607751 ``` { "dim": 352, "cond_dim": 512, "image_embed_dim": 768, "text_embed_dim": 768, "cond_on_text_encodings": true, "channels": 3, "dim_mults": [1, 2, 3, 4],...

let's compare with https://github.com/lucidrains/DALLE2-pytorch/issues/27#issuecomment-1179367413

looks like they use dropout, maybe we should too

nevermind I had batch size at 20 when putting it at 1 I can find Number of parameters: 2297405111 ``` "decoder": { "unets": [ { "dim": 440, "cond_dim": 512, "image_embed_dim":...

save_every_n_samples": 2000000 and epoch_samples": 10000000 are not super convenient because what would be much better is "sample every N minutes" and "evaluate every N minutes" like every hour is good...

> @rom1504 what is the `.not` file extension? Ah that's a typo, i meant npy. Numpy saving format.

Interesting points about the networks that can be trained separately. Will check that out!

I believe we got almost everything now. Time to scale up