Scour VS BLOOM

Scour Reviews

BLOOM Reviews

Scour mentions (6)

• Is it possible to save multiple files as optimized SVGs at once?

  You could install the optimizer Inkscape uses internally and process your files with a super simple shell script. When reading the inkscape command line help, it does not seem to offer any option to export optimized SVGs from there. Source: 11 months ago

• Convert pdf to optimized SVG from command line [Help]

  Inkscape uses Scour under the hood for optimized SVG export. Source: about 1 year ago

• Batch convert svg icons into a symbol set?

  First make sure all your icons are saved as 'Optimized SVG' to remove Inkscape specific data and unnecessary id-attributes. Inkscape uses 'Scour' for this under the hood, so you can just use that directly to convert your files from the command line. I recommend the options --strip-xml-prolog --remove-metadata --enable-id-stripping --renderer-workaround. Now you only need to replace the outermost ... With ... For... Source: almost 2 years ago

• Open Source Favicon Generation & Optimisation in 2022

  The script optimised the SVG using Scour. This removes some metadata and also shortens IDs as well as strip out comments. For the PNG files we used OptiPNG on the maximum optimisation setting. This can be slow on larger files, but for favicons should not take long. Here’s the before and after comparison of files sizes for a particular favicon, using the script:. - Source: dev.to / almost 2 years ago

• Where to start?

  I also recommend Inkscape, it also has scour built-in to it. When saving a SVG select the option "optimized SVG" which will give you options to pass to scour to lower the amount of markup there is. You can then do some hand editing after this to further remove any markup you don't want. Source: over 2 years ago

BLOOM mentions (5)

• How feasible would it be to build a pc for higher parameter models like (quantized) Bloom 176b?

  According to https://huggingface.co/bigscience/bloom it has 70 layers, so that's not a perfect split, but even so that should still fit. Source: 10 months ago

• Do Foundation Model Providers Comply with the EU AI Act?

  It's notable that Hugging Face's BLOOM (https://huggingface.co/bigscience/bloom) might already be compliant (ignoring the 'member states' requirement which I'm sure they could comply with easily enough, it's about disclosing EU member where the model is on the market, so them simply listing all EU countries in a doc somewhere may suffice). Some may be tempted to point at the lower scores of OpenAI, Google and... - Source: Hacker News / 10 months ago

• Ask HN: Where to start in playing with open source generative AI models?

  I've been reading a lot of the latest posts on hacker news, but a little lost on where to actually start. I'd like to run something on my local macbook pro to play around and does anyone have recommendation of like a top 5 links of who to follow and what to play with? So far I've seen things like https://huggingface.co/bigscience/bloom. - Source: Hacker News / 11 months ago

• The greatest risk of AI is from the people who control it, not the tech itself

  They do, and they did. It's called BLOOM and it comes in sizes up to 176B parameters. See: https://huggingface.co/bigscience/bloom. - Source: Hacker News / 11 months ago

• [D] Layers of neurons in LLMs?

  ChatGPT is a GPT-3.5 model according to OpenAI Docs. So it has 175 billion parameters. For the architecture, people said BLOOM is likely the nearest alternative to GPT-3/3.5 models. Source: 12 months ago