min-dalle

AI

text to image model

SHOW HN2022

Legend

min-dalle

I stripped DALL·E Mini to its bare essentials and converted it to Torch

github.com

#dall-e mini#text to image#pytorch#jax#image generation

663points
148comments
#27of 2,570 in 2022
Top 2%of 2022 by points

How HN took it

CuriousMostly questions and neutral comments

7 positive, 37 neutral or asking, 1 mixed, 27 critical, out of 72 comments about the project. Thread score -0.24.

Execution: 2 for, 20 againstIdea: 4 for, 2 againstDocs: 0 for, 3 againstPrivacy: 0 for, 2 againstPerformance: 1 for, 0 against

The thread, comment by comment

Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.

positive neutral or mixed critical about another commenter or something else the maker

What HN said

What people liked

  • Easier to install and run than the JAX version

    From 2 comments
    • Look how much easier it is to install & run, people are interested in and up-voting the result, not how much work was (or wasn't) required to achieve it.
      OJFord on HN
    • Love this, before I only ever saw code that ran this model through jax. This seems to perform much better on my m1.
      jwitthuhn on HN
  • Runs faster on Apple M1 than JAX code

    From 1 comment
    • Love this, before I only ever saw code that ran this model through jax. This seems to perform much better on my m1.
      jwitthuhn on HN

Objections

  • Setup errors with JAX, flax and msgpack versions

    From 3 comments
    • I spun up an aws ubuntu ec2 with 2 Tesla M60. When I run python3 image_from_text.py --text='alien life' --seed=7 I get this error detokenizing image Traceback (most recent call…
      dalle-world on HN
    • Running the "alien life" example from the README took 30 seconds on my M1 Max. I don't think it uses the GPU at all. I couldn't get the "mega" option to work, I got an error…
      tomduncalf on HN
    • Change the flax version to 0.4.2 (currently 0.5.2) will work So much for semver :(
      NhanH on HN
  • Requires a Weights & Biases account to download models

    From 2 comments
    • NB: Needs a weights & biases account in order to download the models.
      pja on HN
    • What is wandb.ai, and Why does it keep asking for an API key? It's not listed in the requirements I've posted it as an issue
      mikewarot on HN
  • Image quality is poor for some prompts

    From 2 comments
    • The results are amazingly poor. Try "biden plays chess against napoleon"
      sAbakumoff on HN
    • Interesting when testing with inputs like "Oscar Wilde photo" or "marilyn monroe photo" and comparing to a Google image search. After some iterations we can have quite similar…
      JacobiX on HN
  • Slow inference on CPU, about 30 seconds or more

    From 2 comments
    • Running the "alien life" example from the README took 30 seconds on my M1 Max. I don't think it uses the GPU at all. I couldn't get the "mega" option to work, I got an error…
      tomduncalf on HN
    • The thread now has a fix. As for the GPU, it's possible to get it working with some extra steps https://github.com/google/jax/issues/8074 Macbook Pro M1 Pro numbers (CPU): ```…
      klohto on HN

Requests

  • Hardware and RAM requirements and inference stats

    From 2 comments
  • Docker container support

    From 1 comment
    • Now we just need to containarise it (there are a few docker python nvidia images)
      lastdong on HN
  • GPU support on Apple silicon

    From 1 comment
    • From reading that thread it didn't sound like GPU was fully supported yet, were you able to get it working?
      tomduncalf on HN

Questions

  • What is Weights & Biases and why is an API key needed

    From 2 comments
    • What is wandb.ai, and Why does it keep asking for an API key? It's not listed in the requirements I've posted it as an issue
      mikewarot on HN
    • WTF is Wandb.ai? This seems like a sneaky way to get people to sign up for this wandb thingy.
      mahastore on HN
  • What is the license of generated images

    From 1 comment
    • What is the license of the generated artwork?
      JanSt on HN

Alternatives named

  • Visions of Chaos for running text-to-image models locally on Windows

    From 1 comment
    • If people just want to run text to image models locally by far the easiest way I've found on Windows is to install Visions of Chaos. It was originally a fractal image generation…
      andybak on HN
  • dalle-playground

    From 1 comment
    • People might also like this one: https://github.com/saharmor/dalle-playground Really easy to work with
      amar-laksh on HN

Category, keywords, summaries and labels on this page were made by AI (Thread Reader v2) and not checked by hand. Each point links to the comments it came from.

Add a badge to your READMEShown on HackerShow
[![Shown on HackerShow](https://hackershow.org/badge.svg)](https://hackershow.org/show/31903076)