We bought circa 10 SOL today (worked out at 9.8 in the end) so that we can start uploading the images on mainnet for our first batches. Our aim is to release 250 NFT’s for Jack and hopefully another 250 from Katie over the coming days which will be a Read more…
The Jack Candy Machine We have been making good progress on Katie and Jack, we have candidates for release and are now updating the site with Jack images. So, we have deployed a Candy Machine for Jack on the Solana devnet. We are currently testing its functionality in addition to Read more…
Mark Adkins the MD has decided to start pushing social media so he is registering all the various platforms. We should be ready to start selling in February as such we need to start making noise about the project. The beginning of this year will be focused on Jack and Read more…
We are taking the photographs from SGD Photos and setting the A100’s to work to create artwork from the photographs. The learning model will in-take artistic elements in addition to photographs that have been processed to produce abstract artwork. This comes from the Cain model which to this date has Read more…
Results from Imogen are good but this was from a very limited dataset. We are going to have the paintings re-photographed and re-worked then prepared for input. On completion, we will set the model learning and hopefully, we should be able to see some results within a few weeks. The Read more…
Well the A100 series did the trick, we can now generate at 1024×1024 – Yes, this takes a long time, but we have been running the models for weeks now. So far, the samples have been very interesting and we are classifying them as a success even before the model Read more…
We agreed to take an A100 hypervisor from Coreix which are the most powerful GPU’s on the market. This has come at quite an expense but we are able to run test models on our other hypervisors while we use these for production generation. We are looking to start releasing Read more…
We deployed a new hypervisor with the V100 cards which have 32GB of memory. Whilst the model took some time we were able to generate 256×256 images with an adapted Gillian model. We are debating looking at the A100 GPU’s since the speed of these is far beyond the V100 Read more…
Gillian worked but not to the resolution we require sadly. We have optimised as much as we can, split it over the GPUs and tried to multi-process but at the end of the day, we just need more GPU memory. I think we have accepted that now. There is a Read more…
We continued from the success of Daniella, modified the model and this model was able to generate 128×128 images. The learning is taking quite some time i.e. to generate models, however, we are seeing a lot of success now. We are going to push ahead with the next generation which Read more…