81 comments

  • fg1371 day ago
    How does this compare to Andrej Karpathy&#x27;s microgpt (<a href="https:&#x2F;&#x2F;karpathy.github.io&#x2F;2026&#x2F;02&#x2F;12&#x2F;microgpt&#x2F;" rel="nofollow">https:&#x2F;&#x2F;karpathy.github.io&#x2F;2026&#x2F;02&#x2F;12&#x2F;microgpt&#x2F;</a>) or minGPT (<a href="https:&#x2F;&#x2F;github.com&#x2F;karpathy&#x2F;minGPT" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;karpathy&#x2F;minGPT</a>)?
    • armanified1 day ago
      I haven&#x27;t compared it with anything yet. Thanks for the suggestion; I&#x27;ll look into these.
    • BrokenCogs1 day ago
      Who cares how it compares, it&#x27;s not a product it&#x27;s a cool project
      • tantalor1 day ago
        Even cool projects can learn from others. Maybe they missed something that could benefit the project, or made some interesting technical choice that gives a different result.<p>For the readers&#x2F;learners, it&#x27;s useful to understand the differences so we know what details matter, and which are just stylistic choices.<p>This isn&#x27;t art; it&#x27;s science &amp; engineering.
        • BrokenCogs1 day ago
          But it isn&#x27;t the OP&#x27;s responsibility to compare their project to all other projects. The GP could themselves perform the comparison and post their thoughts instead of asking an open ended question.
          • philipallstar1 day ago
            &gt; it isn&#x27;t the OP&#x27;s responsibility to compare their project to all other projects<p>No one, including the GP, said it was.
          • fg1371 day ago
            It isn&#x27;t, but such information will be immensely helpful to anyone who wants to learn from such projects. Some tutorials are objectively better than others, and learners can benefit from such information.
          • tantalor1 day ago
            100% agree, I didn&#x27;t mean to imply that OP is responsible for that, or that the (lack of) comparison detracts in any way from the work.
      • stronglikedan1 day ago
        &gt; Who cares how it compares<p>Well, the person who asked the question, for one. I&#x27;m sure they&#x27;re not the only one. Best not to assume why people are asking though, so you can save time by not writing irrelevant comments.
      • layer81 day ago
        Microgpt isn’t a product either. Are you saying that differences between cool projects aren’t worth thinking and conversing about?
  • thomasfl1 day ago
    Is there some documentation for this? The code is probably the simplest (Not So) Large Language Model implementation possible, but it is not straight forward to understand for developers not familiar with multi-head attention, ReLU FFN, LayerNorm and learned positional embeddings.<p>This projects shares similarities with Minix. Minix is still used at universities as an educational tool for teaching operating system design. Minix is the operating system that taught Linus Torvalds how to design (monolithic) operating systems. Similarly having students adding capabilities to GuppyLM is a good way to learn LLM design.
    • achenatx1 day ago
      give the code to an LLM and have a discussion about it.
      • dominotw1 day ago
        does this work? there is no more need for writing high level docs?
        • &gt; does this work?<p>Absolutely. If you loaded this into an agentic coding harness with a decent model, I can practically guarantee it would be able to help you figure out what&#x27;s going on.<p>&gt; there is no more need for writing high level docs?<p>Absolutely not. That would be like exploring a cave without a flashlight, knowing that you could just feel your way around in the dark instead.<p>Code is not always self-documenting, and can often tell you how it was written, but not why.
          • stronglikedan1 day ago
            &gt; If you loaded this into an agentic coding harness with a decent model, I can practically guarantee it would be able to help you figure out what&#x27;s going on.<p>My non-coder but technically savvy boss has been doing this lately to great success. It&#x27;s nice because I spend less time on it since the model has taken my place for the most part.
            • libria1 day ago
              &gt; since the model has taken my place for the most part<p>Hah, you realize the same thing is going on in your boss&#x27;s head right? The pie chart of Things-I-Need-stronglikedan-For just shrank tiny bit...
              • dominotw1 day ago
                my last employer was using ai to rank developers on most impactful code their prs are shipping.
        • sigmoid101 day ago
          There are so many blogs and tutorials about this stuff in particular, I wouldn&#x27;t worry about it being outside the training data distribution for modern LLMs. If you have a scarce topic in some obscure language I&#x27;d be more careful when learning from LLMs.
        • bigmadshoe1 day ago
          LLMs can tell you what the code does but not why the developer chose to do it that way.<p>Also, large codebases are harder to understand. But projects like these are simple to discuss with an LLM.
          • stronglikedan1 day ago
            &gt; LLMs can tell you what the code does but not why the developer chose to do it that way.<p>Do LLMs not take comments into consideration? (Serious question - I&#x27;m just getting into this stuff)
            • bigmadshoe15 hours ago
              They do. Think of it like a very intelligent but somewhat unreliable engineer you can hire to look at your code. They have no context about the codebase beyond what’s written in the source code, or any docs you give them.<p>What I meant was the docs might provide explanations about the problems the codebase solves, design decisions, the abstractions chosen, etc that wouldn’t live in a particular source file. Any discussion someone has with an LLM about the codebase will lack this context in the explanations given if docs don’t exist.
            • dr_hooo1 day ago
              They do (it&#x27;s just text), if they are there...
    • BorisMelnik17 hours ago
      I haven&#x27;t heard minix in so long!
  • totetsu1 day ago
    <a href="https:&#x2F;&#x2F;bbycroft.net&#x2F;llm" rel="nofollow">https:&#x2F;&#x2F;bbycroft.net&#x2F;llm</a> has 3d Visualization of tiny example LLM layers that do a very good job at showing what is going on (<a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=38505211">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=38505211</a>)
    • armanified1 day ago
      Pretty neat! I&#x27;ll definitely take a deeper look into this.
    • devsteru22 hours ago
      Thanks for sharing
    • maverickxone1 day ago
      have little to do with this, but i have to say your project are indeed pretty cool! Consider adding some more UI?
    • skramzy1 day ago
      Neat!
  • ordinarily2 days ago
    It&#x27;s genuinely a great introduction to LLMs. I built my own awhile ago based off Milton&#x27;s Paradise Lost: <a href="https:&#x2F;&#x2F;www.wvrk.org&#x2F;works&#x2F;milton" rel="nofollow">https:&#x2F;&#x2F;www.wvrk.org&#x2F;works&#x2F;milton</a>
  • mudkipdev2 days ago
    This is probably a consequence of the training data being fully lowercase:<p>You&gt; hello Guppy&gt; hi. did you bring micro pellets.<p>You&gt; HELLO Guppy&gt; i don&#x27;t know what it means but it&#x27;s mine.
    • functional_dev2 days ago
      Great find! It appears uppercase tokens are completely unknonw to the tokenizer.<p>But the character still comes through in response :)
  • algoth11 day ago
    This really makes me think if it would be feasible to make an llm trained exclusively on toki pona (<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Toki_Pona" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Toki_Pona</a>)
    • MarkusQ1 day ago
      There isn&#x27;t enough training data though, is there? The &quot;secret sauce&quot; of LLMs is the vast amount of training data available + the compute to process it all.
      • algoth11 day ago
        I think you could probably feed a copy of a toki pona grammar book to a big model, and have it produce ‘infinite’ training data
        • MarkusQ1 day ago
          This is essentially a distillation on the bigger model; you&#x27;d wind up surfacing a lot of artifacts from the host model, amplifying them in the same way repeated photocopying introduces errors.<p><a href="https:&#x2F;&#x2F;dailyai.com&#x2F;2025&#x2F;05&#x2F;create-a-replica-of-this-image-dont-change-anything-ai-trend-takes-off&#x2F;" rel="nofollow">https:&#x2F;&#x2F;dailyai.com&#x2F;2025&#x2F;05&#x2F;create-a-replica-of-this-image-d...</a>
        • eden-u41 day ago
          There are not enough samples in that book to generate new &quot;infinite&quot; data.
    • mudkipdev1 day ago
      People have made toki pona translation models before, not exclusively trained though
  • neurworlds1 day ago
    Cool project. I&#x27;m working on something where multiple LLM agents share a world and interact with each other autonomously. One thing that surprised me is how much the &quot;world&quot; matters — same model, same prompt, but put it in a system with resource constraints, other agents, and persistent memory, the behavior changes dramatically. Made me realize we spend too much time optimizing the model and not enough thinking about the environment it operates in.
  • SilentM682 days ago
    Would have been funny if it were called &quot;DORY&quot; due to memory recall issues of the fish vs LLMs similar recall issues :)
    • armanified1 day ago
      OMG! Why didn&#x27;t I thought fo this first :P
  • zwaps2 days ago
    I like the idea, just that the examples are reproduced from the training data set.<p>How does it handle unknown queries?
    • armanified1 day ago
      It mostly doesn&#x27;t, at 9M it has very limited capacity. The whole idea of this project is to demonstrate how Language Models work.
  • brcmthrowaway2 days ago
    Why are there so many dead comments from new accounts?
    • 59nadir1 day ago
      Because despite what HN users seem to think, HN is a LLM-infested hellscape to the same degree as Reddit, if not more.
      • wiseowise1 day ago
        You’re absolutely right! HN isn’t just LLM-infested hellscape, it’s a completely new paradigm of machine assisted chocolate-infused information generation.
        • toyg1 day ago
          Just let me know which type of information goo you&#x27;d like me to generate, and I&#x27;ll tailor the perfect one for you.
      • siva71 day ago
        But what should we do? The parent company isn&#x27;t transparent about communicating the seriousness of this problem
    • loveparade2 days ago
      It really seems it&#x27;s mostly AI comments on this. Maybe this topic is attractive to all the bots.
      • armanified1 day ago
        This title might have triggered something in those bots; most of them have sneaky AI SaaS links in their bio.<p>Honestly, I never expected this post to become so popular. It was just the outcome of a weekend practice session.
    • AlecSchueler2 days ago
      They all seem to be slop comments.
  • bblb1 day ago
    Could it be possible to train LLM only through the chat messages without any other data or input?<p>If Guppy doesn&#x27;t know regular expressions yet, could I teach it to it just by conversation? It&#x27;s a fish so it wouldn&#x27;t probably understand much about my blabbing, but would be interesting to give it a try.<p>Or is there some hard architectural limit in the current LLM&#x27;s, that the training needs to be done offline and with fairly large training set.
    • roetlich1 day ago
      What does &quot;done offline&quot; mean? Otherwise you are limited by context window.
    • tatrions1 day ago
      [flagged]
  • AndrewKemendo2 days ago
    I love these kinds of educational implementations.<p>I want to really praise the (unintentional?) nod to Nagel, by limiting capabilities to representation of a fish, the user is immediately able to understand the constraints. It can only talk like a fish cause it’s very simple<p>Especially compared to public models, thats a really simple correspondence to grok intuitively (small LLM &gt; only as verbose as a fish, larger LLM &gt; more verbose) so kudos to the author for making that simple and fun.
    • dvt2 days ago
      &gt; the user is immediately able to understand the constraints<p>Nagel&#x27;s point was quite literally the <i>opposite</i>[1] of this, though. We can&#x27;t understand what it must &quot;be like to be a bat&quot; because their mental model is so fundamentally different than ours. So using all the human language tokens in the world can&#x27;t get us to truly understand what it&#x27;s like to be a bat, or a guppy, or whatever. In fact, Nagel&#x27;s point is arguably even stronger: there&#x27;s no <i>possible</i> mental mapping between the experience of a bat and the experience of a human.<p>[1] <a href="https:&#x2F;&#x2F;www.sas.upenn.edu&#x2F;~cavitch&#x2F;pdf-library&#x2F;Nagel_Bat.pdf" rel="nofollow">https:&#x2F;&#x2F;www.sas.upenn.edu&#x2F;~cavitch&#x2F;pdf-library&#x2F;Nagel_Bat.pdf</a>
      • Terr_2 days ago
        IMO we&#x27;re a step before that: We don&#x27;t even have a <i>real</i> fish involved, we have a <i>character</i> that is fictionally a fish.<p>In LLM-discussions, obviously-fictional characters can be useful for this, like if someone builds a &quot;Chat with Count Dracula&quot; app. To truly believe that a typical &quot;AI&quot; is some entity that &quot;wants to be helpful&quot; is just as mistaken as believing the same architecture creates an entity that &quot;feels the dark thirst for the blood of the living.&quot;<p>Or, in this case, that it really enjoys food-pellets.
      • andoando2 days ago
        Id highly disagree with that. Were all living in the same shared universe, and underlying every intelligence must be precisely an understanding of events happening in this space-time.
        • vixen991 day ago
          What does &#x27;precisely&#x27; mean? Everyone has the same understanding of events - a precise one?
          • andoando1 day ago
            No I am saying the basis of intelligence must be shared, not that we have the same exact mental model.<p>I might for example say a human entered a building, a bat might on the other hand think &quot;some big block with two sticks moved through a hole&quot;, but both are experiencing a shared physical observation, and there is some mapping between the two.<p>Its like when people say, if there are aliens they would find the same mathematical constants thet we do
      • AndrewKemendo2 days ago
        Different argument<p>I’m not going to argue other than to say that you need to view the point from a third party perspective evaluating “fish” vs “more verbose thing,” such that the composition is the determinant of the complexity of interaction (which has a unique qualia per nagel)<p>Hence why it’s a “unintentional nod” not an instantiation
  • cbdevidal2 days ago
    &gt; you&#x27;re my favorite big shape. my mouth are happy when you&#x27;re here.<p>Laughed loudly :-D
    • vunderba2 days ago
      This is a direct output from the synthetic training data though - wonder if there is a bit of overfitting going on or it’s just a natural limitation of a much smaller model.
  • rpdaiml1 day ago
    This is a nice idea. A tiny implementation can be way more useful for learning than yet another wrapper around a big model, especially if it keeps the training loop and inference path small enough to read end to end.
  • BiraIgnacio1 day ago
    Nice work and thanks for sharing it!<p>Now, I ask, have LLMs ben demystified to you? :D<p>I am still impressed how much (for the most part) trivial statistics and a lot of compute can do.
  • jzer0cool1 day ago
    Does this work by just training once with next token prediction? Want to understand better how it creates fluent sentences if anyone can provide insights.
  • kaipereira2 days ago
    This is so cool! I&#x27;d love to see a write-up on how made it, and what you referenced because designing neural networks always feel like a maze ;)
  • Leomuck1 day ago
    Wow that is such a cool idea! And honestly very much needed. LLMs seem to be this blackbox nobody understands. So I love every effort to make that whole thing less mysterious. I will definitely have a look at dabbling with this, may it not be a goldfish LLM :)
  • ankitsanghi2 days ago
    Love it! I think it&#x27;s important to understand how the tools we use (and will only increasingly use) work under the hood.
  • NyxVox2 days ago
    Hm, I can actually try the training on my GPU. One of the things I want to try next. Maybe a bit more complex than a fish :)
  • CaseFlatline1 day ago
    I am trying to find how the synthetic data was created (looking through the repo) and didn&#x27;t find it. Maybe I am missing it - Would love to see the prompts and process on that aspect of the training data generation!
    • vunderba1 day ago
      It&#x27;s here:<p><a href="https:&#x2F;&#x2F;github.com&#x2F;arman-bd&#x2F;guppylm&#x2F;blob&#x2F;main&#x2F;guppylm&#x2F;generate_data.py" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;arman-bd&#x2F;guppylm&#x2F;blob&#x2F;main&#x2F;guppylm&#x2F;genera...</a><p>Uses a sort of mad-libs templatized style to generate all the permutations.
  • ergocoder1 day ago
    It&#x27;s just so amazing that 5 years ago it would be extremely to build a conversational bot like this.<p>But right now people make it a hobby, and that thing can run on a laptop.<p>This is just so wild.
  • Duplicake1 day ago
    I love this! Seems like it can&#x27;t understand uppercase letters though
    • armanified1 day ago
      Uppercase letters were intentionally ignored.
  • bharat10101 day ago
    This is such a smart way to demystify LLMs. I really like that GuppyLM makes the whole pipeline feel approachable..great work
  • drincanngao1 day ago
    I was going to suggest implementing RoPE to fix the context limit, but realized that would make it anatomically incorrect.
    • armanified1 day ago
      I intentionally removed all optimizations to keep it vanilla.
  • gnarlouse2 days ago
    I... wow, you made an LLM that can actually tell jokes?
    • murkt2 days ago
      With 9M params it just repeats the joke from a training dataset.
  • kubrador2 days ago
    how&#x27;s it handle longer context or does it start hallucinating after like 2 sentences? curious what the ceiling is before the 9M params
  • fawabc1 day ago
    how did you generate the synthetic data?
  • Building it yourself is always the best test if you really understand how it works.
  • amelius1 day ago
    &gt; A 9M model can&#x27;t conditionally follow instructions<p>How many parameters would you need for that?
    • armanified1 day ago
      My initial idea was to train a navigation decision model with 25M parameters for a Raspberry Pi, which, in testing, was getting about 60% of tool calls correct. IMO, it seems like around 20M parameters would be a good size for following some narrow &amp; basic language instructions.
      • amelius1 day ago
        Ok. This makes me wonder about a broader question. Is there a scientific approach showing a pyramid of cognitive functions, and how many parameters are (minimally) required for each layer in this pyramid?
  • ananandreas1 day ago
    Great and simple way to bridge the gap between LLMs and users coming in to the field!
  • rclkrtrzckr2 days ago
    I could fork it and create TrumpLM. Not a big leap, I suppose.
    • probably 8M params are too much even :)
      • danparsonson1 day ago
        As long as you use the best parameters then it doesn&#x27;t matter
      • wiseowise1 day ago
        Grab her by the pointer.
  • Thanks. Tinkering is how I learn and this is what I’ve been looking for.
  • jbethune1 day ago
    Forked. Very cool. I appreciate the simplicity and documentation.
  • ben8bit1 day ago
    This is really great! I&#x27;ve been wanting to do something similar for a while.
  • tombelieber17 hours ago
    looking forward to try it, great job
  • nullbyte8082 days ago
    Adorable! Maybe a personality that speaks in emojis?
    • armanified1 day ago
      OMG! You just gave me the next idea..
  • winter_blue1 day ago
    This is amazing work. Thank you.
  • monksy2 days ago
    Is this a reference from the Bobiverse?
  • cpldcpu2 days ago
    Love it! Great idea for the dataset.
  • * How creating dataset? I download it but it is commpresed in binary format.<p>* How training. In cloud or in my own dev<p>* How creating a gguf
    • ``` uv run python -m guppylm chat<p>Traceback (most recent call last):<p><pre><code> File &quot;&lt;frozen runpy&gt;&quot;, line 198, in _run_module_as_main File &quot;&lt;frozen runpy&gt;&quot;, line 88, in _run_code File &quot;&#x2F;home&#x2F;user&#x2F;gupik&#x2F;guppylm&#x2F;guppylm&#x2F;__main__.py&quot;, line 48, in &lt;module&gt; main() File &quot;&#x2F;home&#x2F;user&#x2F;gupik&#x2F;guppylm&#x2F;guppylm&#x2F;__main__.py&quot;, line 29, in main engine = GuppyInference(&quot;checkpoints&#x2F;best_model.pt&quot;, &quot;data&#x2F;tokenizer.json&quot;) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File &quot;&#x2F;home&#x2F;user&#x2F;gupik&#x2F;guppylm&#x2F;guppylm&#x2F;inference.py&quot;, line 17, in __init__ self.tokenizer = Tokenizer.from_file(tokenizer_path) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^</code></pre> Exception: No such file or directory (os error 2) ```
      • meybe add training again (read best od fine) and train again<p>``` # after config device checkpoint_path = &quot;checkpoints&#x2F;best_model.pt&quot;<p>ckpt = torch.load(checkpoint_path, map_location=device, weights_only=False)<p>model = GuppyLM(mc).to(device) if &quot;model_state_dict&quot; in ckpt: model.load_state_dict(ckpt[&quot;model_state_dict&quot;]) else: model.load_state_dict(ckpt)<p>start_step = ckpt.get(&quot;step&quot;, 0) print(f&quot;Encore {start_step}&quot;) ```
    • freetonik1 day ago
      You sound like Guppy. Nice touch.
  • rahen1 day ago
    I don&#x27;t mean to be &#x27;that guy&#x27;, but after a quick review, this really feels like low-effort AI slop to me.<p>There is nothing wrong using AI tools to write code, but nothing here seems to have taken more than a generic &#x27;write me a small LLM in PyTorch&#x27; prompt, or any specific human understanding.<p>The bar for what constitutes an engineering feat on HN seems to have shifted significantly.
    • zhainya7 hours ago
      I don&#x27;t really understand the point of this project or how it demystifies anything. Click the browser demo and I get a generic AI chat screen. Is the readme the part that &quot;demystifies&quot; something? I feel like I am living in a bizarro world. Is this all AI? Are all the comments here from bots?
  • Haha, funny name :)
  • Elengal1 day ago
    Cool
  • oyebenny2 days ago
    Neat!
  • hughw1 day ago
    Tiny LLM is an oxymoron, just sayin.
    • uxcolumbo1 day ago
      How about: LLMs are on a spectrum and this one is on the tiny side?
    • armanified1 day ago
      True, but most would ignore LM if it weren&#x27;t LLM.
  • hahooh1 day ago
    haha funny, but really cool project. why fish tho lol.
  • hackerman700002 days ago
    [flagged]
    • zkmon1 day ago
      Meaning&#x2F;goal of life is to reproduce. Food (and everything else) is only a means to it. Reproduction is the only root goal given by nature to any life form. All resources and qualities are provided are only to help mating.
      • tantalor1 day ago
        Reproduction is the goal of <i>genes</i>.<p>Food (not dying) is the goal of organisms.
        • philote1 day ago
          I&#x27;d argue genes nor life has a &quot;goal&quot;. They are what they are because they&#x27;ve been successful at continuing their existence. Would you say a rock&#x27;s goal is not to get broken?
          • tantalor1 day ago
            Only because genes&#x2F;organisms can make choices (changes to its programming, or decisions) to optimize their path towards their goal.<p>A rock is maybe not a good counterexample, but a crystal is because it can grow over time. So in some sense, it tries not to break. However a crystal cannot make any choices; it&#x27;s behavior is locked into the chemistry it starts with.
      • hca1 day ago
        No, evolution has encoded lust. It has not yet allowed for condoms. But it&#x27;s a process.
      • amelius1 day ago
        Then why are reproductive rates so low in western countries?<p><a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;List_of_countries_by_total_fertility_rate#&#x2F;media&#x2F;File:Total_Fertility_Rate_Map_by_Country.svg" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;List_of_countries_by_total_fer...</a>
        • darepublic1 day ago
          The western lifestyle is an evolutionary dead end?
          • vixen991 day ago
            It seems that some in the West want it to be and are working hard to make it so.
        • michaelhoney1 day ago
          not just western countries
      • inanutshellus19 hours ago
        I don&#x27;t get why anyone downvoted you but maybe we can &quot;all get along&quot; by saying:<p><pre><code> &quot;the meaning of life is to continue living.&quot; </code></pre> Thus the short-term answer is &quot;food&quot; and &quot;reproduction&quot; is the long-term answer.
    • amelius1 day ago
      It&#x27;s arguably even better than the most famous answer to that question.
      • siva71 day ago
        which is?
        • amelius1 day ago
          <a href="https:&#x2F;&#x2F;medium.com&#x2F;change-your-mind&#x2F;the-meaning-of-life-is-42-9de7a3571312" rel="nofollow">https:&#x2F;&#x2F;medium.com&#x2F;change-your-mind&#x2F;the-meaning-of-life-is-4...</a>
  • meidad_g1 day ago
    [flagged]
  • techpulselab1 day ago
    [dead]
  • textai20261 day ago
    [dead]
  • Morpheus_Matrix2 days ago
    [flagged]
  • adamsilvacons1 day ago
    [dead]
  • maxothex1 day ago
    [dead]
  • solsafe_dev1 day ago
    [dead]
  • ethanmacavoy2 days ago
    [flagged]
  • agdexai1 day ago
    [dead]
  • agenexus2 days ago
    [flagged]
  • Alexzoofficial2 days ago
    [flagged]
  • peifeng072 days ago
    [dead]
  • zephyrwhimsy1 day ago
    [flagged]
  • zephyrwhimsy2 days ago
    [flagged]
  • zephyrwhimsy1 day ago
    [flagged]
  • novachen1 day ago
    [dead]
  • zephyrwhimsy1 day ago
    [flagged]
  • weiyong10242 days ago
    [flagged]
  • aesopturtle2 days ago
    [flagged]
  • techpulselab2 days ago
    [dead]
  • [dead]
  • george_belsky2 days ago
    [dead]
  • solsafe_dev1 day ago
    [dead]
  • aditya73030112 days ago
    Did something similar last year <a href="https:&#x2F;&#x2F;github.com&#x2F;aditya699&#x2F;EduMOE" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;aditya699&#x2F;EduMOE</a>
  • zhichuanxun1 day ago
    [dead]
  • aditya73030112 days ago
    [dead]
  • martmulx2 days ago
    [flagged]
  • LeonTing10102 days ago
    [flagged]
    • secabeen2 days ago
      Training data is here:<p><a href="https:&#x2F;&#x2F;huggingface.co&#x2F;datasets&#x2F;arman-bd&#x2F;guppylm-60k-generic" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;datasets&#x2F;arman-bd&#x2F;guppylm-60k-generic</a>
  • areys1 day ago
    [flagged]
    • moonu1 day ago
      This comment seems ai-written
  • jiusanzhou2 days ago
    [flagged]
    • ngruhn2 days ago
      comment smells AI written
    • 3m2 days ago
      AI account
  • dinkumthinkum2 days ago
    I think this is a nice project because it is end to end and serves its goal well. Good job! It&#x27;s a good example how someone might do something similar for a specific purpose. There are other visualizers that explain different aspects of LLMs but this is a good applied example.
  • Propelloni1 day ago
    Great work! I still think that [1] does a better job of helping us understand how GPT and LLM work, but yours is funnier.<p>Then, some criticism. I probably don&#x27;t get it, but I think the HN headline does your project a disservice. Your project does not demystify anything (see below) and it diverges from your project&#x27;s claim, too. Furthermore, I think you claim too much on your github. &quot;This project exists to show that training your own language model is not magic.&quot; and then just posts a few command line statements to execute. Yeah, running a mail server is not magic, just apt-get install exim4. So, code. Looking at train_guppylm.ipynb and, oh, it&#x27;s PyTorch again. I&#x27;m better off reading [2] if I&#x27;m looking into that (I know, it is a published book, but I maintain my point).<p>So, in short, it does not help the initiated or the uninitiated. For the initiated it needs more detail for it to be useful, the uninitiated more context for it to be understood. Still a fun project, even if oversold.<p>[1] <a href="https:&#x2F;&#x2F;spreadsheets-are-all-you-need.ai&#x2F;" rel="nofollow">https:&#x2F;&#x2F;spreadsheets-are-all-you-need.ai&#x2F;</a> [2] <a href="https:&#x2F;&#x2F;github.com&#x2F;rasbt&#x2F;LLMs-from-scratch" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;rasbt&#x2F;LLMs-from-scratch</a>
    • jadengeller1 day ago
      this comment seems to be astroturfing to sell a course
      • Propelloni11 hours ago
        What do you mean, the LLM from Scratch book?