GX Kok
GX Kok
Here are some of my thoughts. 1. We all want to train the most advanced models and I'm feeling openai has the most advanced models (dalle2) but the only open...
> > > > I think latent diffusion you always run into the problem with the codebook which is already limited so there will be a barrier on the quality...
Perhaps it is best to train 1 model at a time.
@searchivarius Yes, each call to addDataPointBatch() does add more data to the index, the length of the index is correct as shown in the console log after call to CreateIndex()....
I also build from source to better utilize the hardware if that means anything.
These embeddings are image features, created from the same number of files. I have not checked the images for exact duplicates, but there should be similar images. Let me see...
Here is a [link](https://drive.google.com/drive/folders/1YnhbkXiNEkLWt8scT-iDKPz2s1uWhDDW?usp=sharing) to some (4) pickle files containing image embeddings and image ids/list. And the following is a sample Python script to reproduce the error. Put the pickle...
Ok noted, thank you so much for the quick response. Glad to be of help.
I think maybe no need to use GPU for testing. Put CUDA_VISIBLE_DEVICES="" before the test command.
Just changing the --target_size command line argument to 512?