Log inSign up
Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
7,312 posts
user avatar
Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
@morgymcg
Applied AI @wandb / @coreweave | ex-Facebook, electricity trading | github.com/morganmcg1 | ๐Ÿ‡ฎ๐Ÿ‡ช | Came for the bants, stayed for the rants
Ireland
ntentional.com
Joined October 2010
4,940
Following
3,459
Followers

New to X?

Sign up now to get your own personalized timeline!

Create account

By signing up, you agree to the Terms of Service and Privacy Policy, including Cookie Use.

TermsยทPrivacyยทCookiesยทAccessibilityยทAds Infoยทยฉ 2026 X Corp.
Don't miss what's happening
People on X are the first to know.
Log inSign up
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Mar 19, 2021
    ๐Ÿ˜ This @huggingface tip to prevent colab from disconnecting ` function ConnectButton(){ console.log("Connect pushed"); document.querySelector("#top-toolbar > colab-connect-button").shadowRoot.querySelector("#connect").click() } setInterval(ConnectButton,60000); `
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Aug 30, 2021
    The #fastai forums are often a goldmine for ML and python code in general. Great breakdown of parallel processing in python here: forums.fast.ai/t/python-paralโ€ฆ
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Nov 13, 2023
    The Llama 2 Getting Started guide from @AIatMeta is really comprehensive, with plenty of code examples for fine-tuning and inference Delighted to see @wandb added there as the logger of choice ๐Ÿคฉ See the guide from Meta here: ai.meta.com/llama/get-starโ€ฆ From what I can
    98K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Jan 25, 2020
    After a dismal performance in the @kaggle RSNA Intracranial Haemorrhage competition I pulled apart the 2nd place solution to see what they had done so right. Post 3 of 4 now up with jupyter notebooks :D Blog series: ntentional.squarespace.com/technical/2020โ€ฆ Notebooks: github.com/morganmcg1/kagโ€ฆ
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Aug 22, 2020
    I put together a quick fastai demo implementing @karpathy's notebook training minGPT to generate Shakespeare Code: gist.github.com/morganmcg1/b2aโ€ฆ Its a quick demo, follow along on the fastai forum to see how it progresses: forums.fast.ai/t/mingpt-fastaโ€ฆ
    user avatar
    Andrej Karpathy
    @karpathy
    Aug 17, 2020
    I wrote a minimal/educational GPT training library in PyTorch, am calling it minGPT as it is only around ~300 lines of code: github.com/karpathy/minGPT +demos for addition and character-level language model. (quick weekend project, may contain sharp edges)
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Aug 27, 2023
    This talk from @colinraffel at the @SimonsInstitute highlighting the advantages of an ecosystem of specialist models/adapters was great - good starting place if youโ€™re curious about how to combine/hot-swap LLM adapters youtube.com/live/AeRmk-gn8โ€ฆ
    20K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Dec 6, 2023
    Given theres no sign of being able to use Ultra any time soon, this @AnthropicAI post in response to @GregKamradt 's Claude 2.1's evaluation is probably the more useful tweet for your day today ๐Ÿ˜ƒ
    A graph showing Claude 2.1's ability to recall an individual sentence with near complete fidelity across its 200K token context window.
    user avatar
    Anthropic
    @AnthropicAI
    Dec 6, 2023
    Claude 2.1โ€™s 200K token context window is powerful, but requires careful prompting to use effectively. Learn how to get Claude to recall an individual sentence across long documents with high fidelity: anthropic.com/index/claude-2โ€ฆ
    28K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Nov 25, 2023
    Any one else strugging with `gpt-4-1106-preview` ? Using @jxnlco 's Instructor and I'm finding `gpt-4-1106-preview` to be really bad at following instructions - only successfully generating examples for 2 out of 7 examples vs consistently 7 out of 7 for `gpt-4-0613` Same issue
    89K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Jun 18, 2023
    Anyone else notice what Falcon 40b does (and doesnโ€™t) like to say about Abu Dhabi > !falcon tell me something interesting โ€œWould you like me to tell you something interesting about technology or something about Abu Dhabi?โ€ Cool, cool, cool
    265K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Jul 27, 2024
    This is a great blog post, while it points out the strong performers in time to first token and token generation rate, it also highlights other practical considerations when choosing between inference libraries:
    user avatar
    Stas Bekman
    @StasBekman
    Jul 27, 2024
    Inference: This was a useful read comparing vLLM, LMDeploy, MLC-LLM, TensorRT-LLM, and TGI with TTFT and Token generation rate comparisons and ease of use and other practical insights bentoml.com/blog/benchmarkโ€ฆ It's concise and to the point, so time well spent. Thank you,
    17K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Oct 9, 2020
    Once you get done with the latest @huggingface newsletter released today, come on over and check out the first in a 2-part series of how to *comprehensively* set up and train a language model using HuggingFace Datasets and Trainer
    ntentional.com
    Comprehensive Language Model Fine Tuning, Part 1: ๐Ÿค— Datasets library [Updated]
    Get your data ready to train with the ๐Ÿค— Datasets library, plus Datasets implementation tips and tricks
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Feb 8, 2024
    A user yesterday just casually dropped a mobile client for @wandb in a GitHub issue๐Ÿ”ฅ๐Ÿ”ฅ๐Ÿ”ฅ Includes all metrics plus system metrics I think this could be really useful to keep an eye on long-running training runs while you're on the move Long-requested, delivered by
    00:00
    22K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Apr 26, 2023
    00:00
    31K
  • user avatar
    Morgan McGuire @ ICML ๐Ÿ‡ฐ๐Ÿ‡ท
    @morgymcg
    Jan 1, 2024
    Anyone want to suggest some Mixtral 8x7b fine-tuning configs to try? Myself and @capetorch have 8 x H100s (thanks to @CoreWeave ๐Ÿ™‡) for 5 more days and are doing some explorations find a decent Mixtral fine-tuning recipe using axolotl that we can share. What configs should we
    user avatar
    Eric Hartford
    @QuixiAI
    Dec 30, 2023
    Dolphin-2.7-mixtral-8x7b is cooking with these fixes: github.com/huggingface/trโ€ฆ github.com/huggingface/trโ€ฆ Hacked to work with axolotl Trying for a tune of Mixtral competes with @MistralAI's mixtral-instruct Thanks @teknium for the consult and @FernandoNetoAi for the pair
    23K