From 3dc92b19bda5d8e7c13c869d76fd0692d8d780de Mon Sep 17 00:00:00 2001 From: Brett Kuprel Date: Mon, 4 Jul 2022 10:22:45 -0400 Subject: [PATCH] update readme --- README.md | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 681f1c7..662528c 100644 --- a/README.md +++ b/README.md @@ -8,10 +8,11 @@ This is a fast, minimal implementation of Boris Dayma's [DALL·E Mega](https://github.com/borisdayma/dalle-mini). It has been stripped down for inference and converted to PyTorch. The only third party dependencies are numpy, requests, pillow and torch. -It takes -- **32 seconds** to generate a 3x3 grid with a P100 in Colab -- **16 seconds** to generate a 4x4 grid with an A100 on Replicate -- **TBD** to generate a 4x4 grid with an H100 (@NVIDIA?) +To generate a 4x4 grid of DALL·E Mega images it takes +- **1m 29s seconds** with a T4 in Colab +- **48 seconds** with a P100 in Colab +- **16 seconds** with an A100 on Replicate +- **TBD** with an H100 (@NVIDIA?) The flax model and code for converting it to torch can be found [here](https://github.com/kuprel/min-dalle-flax).