sangwulee

7 ms·

1.
by sangwulee 3mo ago
A lot of coffee for sure. Regarding the training cost, it's hard to give a good estimate because we used a shared kubernetes cluster with inference + research workloads.
2.
by sangwulee 1y ago
The highest quality finetuning data was hand curated internally. I would say our post training pipeline is quite similar to SeedDream 2.0 ~ 3.0 series from ByteDance. Similar to them, we use extensive quality filters and internal models to
3.
by sangwulee 1y ago
I actually tried a few experiments in early exploration stages! I trained a small classifier to judge AI vs non-AI images. Use it as a reward model to do small RL / post training experiments. Sadly, was not too successful. We found tha
4.
by sangwulee 1y ago
The architecture is the same so we found that some LoRAs work out-of-the box, but some LoRAs don't. In those cases, I would expect people to re-run their LoRA finetuning with the trainer they've used.
5.
by sangwulee 1y ago
We used two types of datasets for post-training. Supervised finetuning data and preference data used for RLHF stage. You can actually use less than < 1M samples to significantly boost the aesthetics. Quality matters A LOT. Quantity helps
6.
by sangwulee 1y ago
We have not added a separate RTX accelerated version for FLUX.1 Krea, but the model is fully compatible with existing FLUX.1 dev codebase. I don't think we made a separate onnx export for it though. Doing 4~8 bit quantized version with
7.
by sangwulee 1y ago
FLUX.1 is one of the most popular open weights text-to-image models. We distilled Krea-1 to FLUX.1 [dev] model so that the community can adopt it seamlessly into existing ecosystem. Any finetuning code, workflows, etc that was built on top
8.
by sangwulee 1y ago
Quick napkin math assuming bfloat16 format : 1B * 16 bits = 16B bits = 2GB. Since it's a 12B parameter model, you get around ~24GB. Downcasting to bfloat16 from float32 comes with pretty minimal performance degradation, so we uploaded
9.
by sangwulee 1y ago
I love owls. Photorealism was one of the focus areas for training because "AI look" (e.g. plastic skin) was biggest complaint for FLUX.1 model series. Photorealism was achieved with both careful curation of finetuning and preferen
10.
by sangwulee 1y ago
Thank you! Glad you find it helpful. The model is focused on photorealism so it should be able to generate most realistic scenes. Although, I think using 3D engines would be more suitable for typical cases for robotics training since it giv
11.
by sangwulee 1y ago
Hi there, I'm Sangwu Lee, one of the researchers behind this model. I'm happy to answer any questions here. --- I also commented in this other submission: https://news.ycombinator.com/item?id=44748056
12.
by sangwulee 1y ago
Hello HackerNews. My name is Sangwu Lee . I work for Krea and I led the research efforts around the post-training for this model. I'll try to answer any questions you may have, but I recommend you read the technical report I wrote on o
13.
by sangwulee 1y ago
Hi! I'm lead researcher on Krea-1. FLUX.1 Krea is a 12B rectified flow model distilled from Krea-1, designed to be compatible with FLUX architecture. Happy to answer any technical questions :)