Hacker Newsnew | past | comments | ask | show | jobs | submit | forgingahead's commentslogin

Yeah, mind sharing any of the scripts? I looked at the docs briefly, looks like we need to install ALL of nemo to get access to Parakeet? Seems ultra heavy.

You only need the ASR bits -- this is where I got to when I previously looked into running Parakeet:

    # NeMo does not run on 3.13+
    python3.12 -m venv .venv
    source .venv/bin/activate

    git clone https://github.com/NVIDIA/NeMo.git nemo
    cd nemo

    pip install torch torchaudio torchvision --index-url https://download.pytorch.org/whl/cu128
    pip install .[asr]

    deactivate
Then run a transcribe.py script in that venv:

    import os
    import sys
    import nemo.collections.asr as nemo_asr

    model_path = sys.argv[1]
    audio_path = sys.argv[2]

    # Load from a local path...
    asr_model = nemo_asr.models.EncDecRNNTBPEModel.restore_from(restore_path=model_path)

    # Or download from huggingface ('org/model')...
    asr_model = nemo_asr.models.EncDecRNNTBPEModel.from_pretrained(model_name=model_path)

    output = asr_moel.transcribe([audio_path])
    print(output[0])
With that I was able to run the model, but I ran out of memory on my lower-spec laptop. I haven't yet got around to running it on my workstation.

You'll need to modify the python script to process the response and output it in a format you can use.


Wish it was on the web app as well!

The HN peanut gallery remains undefeated

System prompts are fine and all, but how useful is it really when LLMs clearly ignore prompt instructions randomly? I've had this with all the different LLMs, explicitly asking it to not do something works maybe 85-90% of the time. Sometimes they just seem "overloaded", even in a fresh chat session, so like a human would, they get confused and drop random instructions.

Perhaps I'm being mean (haven't read the full presentation) - but the winning team is made up of 2 actresses/artists, 1 social innovator, 1 designer, and 1 astrophysicist?

One of the artist is also an environmental engineer, the other is also a psychologist, the designer is an architect.

But yes, it's a work of design more than a blueprint.


Love to see the Ruby implementations! Thanks for sharing.


Thank you so much for sharing this!

We are using ruby to build a powerful AI toolset in the construction space, and we love how simple all of the SaaS parts are and not reinventing the wheel, but the ruby LLM SDK ecosystem is a bit lagging, so we've written a lot of our own low-level tools.

(btw we are also hiring rubyists https://news.ycombinator.com/item?id=43865448)


Great to see someone pushing Ruby in this space. Not after anything permanent, but if you're ever after "plugging gaps" with contract work, complex Ruby projects is what I enjoy the most.


In college, people would just put a sock on the door....


This never happens in real life....."rewriting software" is the introverted programmer's wet dream because it gives them relevance and the idea of respect. No serious business "rewrites software in something else" once they start to take off.


You don't do it for fun*, but because the rapid development duck-typed dynamic language you used to get to MVP quickly is not the language you need to keep it working under load and a growing feature set.

It's a terrible and difficult transition that makes you question if the first language was really such a good choice after all, although it did get you where you are right now, which is more than you can say for a bunch of companies trying to do everything future proof from day 0

(* well, some people do, but they don't tend to survive)


I can point to plenty of companies that have rewritten products at scale. That said, specifically relevant to the article, I believe Shopify and GitHub continue to run Ruby on Rails.


shopify is hugely invested in react and even full stack react through react router. still rails is great for the backend of course


This sound be the top comment, and the discussion should be around why this is the case (how Congress appropriates money like this), what alternatives there could be, and what are the pros/cons of each different option. Sadly HN is no longer a place where this can happen.


Great article, thanks for taking the time to research and write it all up! Definitely learned some new things from it.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: