Implementation of Imagen, Google's text-to-image neural network, in PyTorch
1–10 of 123 posts
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#2How much would it cost to train something like this? Is there even a good dataset for it?
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#3I wonder if this implementation was trained on the same dataset that trained DALL-E2, would it produce results of comparable quality.
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#4I did the pip installs and installed Cuda.
I changed prompt and ran the sample code. It ran to completion. How do I save the image from trainer.sample?
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#5I wonder if this implementation was trained on the same dataset that trained DALL-E2, would it produce results of comparable quality.
I have to clarify I am visioning DALL-E-2-PORN.
Couldn't you just scrape porn to get copious amount of dataset? Scope would be narrower and thus require less classification and in general it has common themes.
I'm more concerned how expensive it will be to train it on GPU instances. We are looking at A6000s right? That's like $15/hr.
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#6How much would it cost to train something like this? Is there even a good dataset for it?
There is a dataset of 5 billion image-text pairs (laion-5b) scraped by various parties. This can then be filtered and used to train these models. Cost is a bit of an issue but there are orgs that have provided compute for open model training. And Imagen is nice because the text encoder part is already available and doesn't need more training, so it would just be the diffusion model components being trained. I'd guess we'll see a biggish training run starting in a few weeks.
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#7I wonder if this implementation was trained on the same dataset that trained DALL-E2, would it produce results of comparable quality.
I have to clarify I am visioning DALL-E-2-PORN. Couldn't you just scrape porn to get copious amount of dataset? Scope would be narrower and thus require less classification and in general it has common themes. I'm more concerned how expensive it will be to train it on GPU instances. We are looking at A6000s right? That's like $15/hr.
"just"
All the complexity wrapped up into one word.
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#8Earlier quoted context omitted.
I have to clarify I am visioning DALL-E-2-PORN. Couldn't you just scrape porn to get copious amount of dataset? Scope would be narrower and thus require less classification and in general it has common themes. I'm more concerned how expensive it will be to train it on GPU instances. We are looking at A6000s right? That's like $15/hr.
"just" All the complexity wrapped up into one word.
You do realize web scraping has gotten very cheap at scale and easy now too. It's an after thought for me, I'm more concerned with the economics of training, it can't be cheap
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#9What's it going to take to train this on porn? This is something that can be crowdfunded.
Look deepnude is a thing and somebody is making money off it: https://app.deepnude.cc/upload
Re: Implementation of Imagen, Google's text-to-image neural network, in PyTorch
#10Earlier quoted context omitted.
"just" All the complexity wrapped up into one word.
You do realize web scraping has gotten very cheap at scale and easy now too. It's an after thought for me, I'm more concerned with the economics of training, it can't be cheap
Web scrape... what, exactly? How will "web scraping" lead to high quality results?