Knut Jägersberg

KnutJaegersberg

AI & ML interests

NLP, opinion mining, narrative intelligence

Recent Activity

Articles

Organizations

LLMs's profile picture Blog-explorers's profile picture Qwen's profile picture Social Post Explorers's profile picture M4-ai's profile picture Chinese LLMs on Hugging Face's profile picture

KnutJaegersberg's activity

posted an update about 13 hours ago
posted an update 4 days ago
posted an update 3 months ago
view post
Post
1173
appvoid/arco

arco consistently outperforms every sota model below 600m parameters on average

appvoid/arco
posted an update 3 months ago
posted an update 4 months ago
posted an update 4 months ago
posted an update 5 months ago
Reacted to merve's post with 👍 5 months ago
posted an update 5 months ago
view post
Post
638
Unsocial Intelligence: an Investigation of the Assumptions of AGI Discourse

I don't agree with some of the assertions made here, but it is an interesting paper and a good overview.

https://arxiv.org/abs/2401.13142
Reacted to merve's post with ❤️ 5 months ago
view post
Post
4325
Florence-2 is a new vision foundation model capable of a wide variety of tasks 🤯
Demo 👉🏻 gokaygokay/Florence-2
Collection 👉🏻 microsoft/florence-6669f44df0d87d9c3bfb76de

This model can handle tasks that vary from OCR to semantic segmentation.

The difference from previous models is that the authors have compiled a dataset consisting of 126M images with 5.4B annotations labelled with their own data engine pseudolabelled by smaller specialized models and APIs.

The model has a similar architecture to previous models: an image encoder and a multimodality encoder with a text decoder. The authors have compiled the multitask dataset with prompts for each task.

You can also fine-tune this model on any task of choice. The authors also released different results on downstream tasks and reported their results when un/freezing the vision encoder 🤓📉
They have released fine-tuned models too, you can find them in the collection above 🤗
·
Reacted to merve's post with 🔥 6 months ago
view post
Post
3011
Finally @CVPR2024 is here! 🩷
Have you claimed your papers and linked your models/datasets/demos?
This will increase visibility and impact of your paper 💫

To index your papers, go here
CVPR2024/CVPR2024-papers
Find your paper, click on paper page link, index the paper, then click on your name (workflow is below 👇🏻)
If you'd like to add links to your paper, go here CVPR2024/update-CVPR2024-papers
login, find your paper's id, retrieve the paper, fill in the info and submit!
replied to s3nh's post 6 months ago
view reply

Don't burn out! Lighten up again will you.

posted an update 6 months ago
Reacted to s3nh's post with ❤️ 6 months ago
view post
Post
GPU Poor POV: Burnout

Sometimes we do not have an energy to post about AI and new methods.
And thats totally ok, I guess.
Remember to sleep well and drink a lot of water. Have a great day :D <3
  • 2 replies
·
replied to BramVanroy's post 8 months ago
view reply

it mixed up stuff in the output, gave weird answers. didn't have that problem with other models. maybe the update they released sovled that issue, I just never cared, given the alternatives.

Reacted to BramVanroy's post with 👍 8 months ago
view post
Post
2391
Does anyone have experience with finetuning Gemma? Even the 2B variant feels more memory heavy than mistral 7B. I know that its vocabulary is much larger (250k) but I'm a bit surprised that the max batch size that I can get in an A100 80GB is only 2 whereas I could fit 4 with mistral 7B - even though Gemma is much smaller except for the embedding layer. Both runs were using FA, same sequence length, same deepspeed zero 3 settings. Oh and yes I'm using the most recent hot fix of transformers that solves a memory issue with Gemma and others.

Any prior experience that you can share or suggestions to improve throughout?
  • 4 replies
·
replied to BramVanroy's post 8 months ago
view reply

I got some weird results, since there are a lot of other models in that performance-parameter range, I just didn't try anymore.

Reacted to clefourrier's post with ❤️ 9 months ago
view post
Post
🔥 New LLM leaderboard blog: Open Ko LLM!

One of the oldest leaderboards on the hub, it has already evaluated more than 1000 models! It uses Korean translations of MMLU, ARC, HellaSwag, TruthfulQA, and a new dataset, Korean CommonGen, about specific common sense alignement.

upstage/open-ko-llm-leaderboard

What's interesting about this leaderboard is how it drove LLM development in Korea, with on average about 4 submissions/models per day since it started!
Really looking forward to seeing similar initiatives in other languages, to help qualitative models emerge outside of "just English" (for the other 2/3rds of the world).

Read more about how the leaderboard in the intro blog: https://huggingface.co/blog/leaderboards-on-the-hub-upstage
Congrats to @Chanjun , @hunkim and the Upstage team!
Reacted to macadeliccc's post with ❤️ 10 months ago
view post
Post
Reducing perplexity in LLM's through layer selective rank reduction

Layer-Selective Rank Reduction (LASER) is a denoising method that improves reasoning by the strategic removal of higher-order components from weight matrices in the multi-layer perceptron (MLP) layers without the need for additional parameters or training data. This process leverages singular value decomposition to identify and eliminate these components. This simple, yet effective, method has shown to improve question-answering performance by up to 27.4 percentage points.

LaserRMT implements this through a process by calculating signal to noise ratio (SNR) for each layer and selectively reducing the rank of these layers.The SNR method meticulously computes the SNR by leveraging singular value decomposition (SVD) to separate the signal (higher-order components) from the noise (lower-order components) within the weight matrices of the model's layers. The SNR calculation is what determines which layers would benefit from rank reduction without compromising the models integrity.

If a layer is identified that could benefit from rank reduction, then the layer will enter an incremental process where the weight matrices are reduced and reconstructed by retaining only the singular values that surpass the threshold. In the case of laserRMT, the threshold is calculated by Marchenko-Pastur Law.
@staticmethod
    def marchenko_pastur_threshold(sigma, n, m):
        beta = n / m if n < m else m / n
        threshold = sigma * np.sqrt((1 + np.sqrt(beta))**2)
        return thr

The two primary benefits of applying this method are reducing computational overhead of large language models and simultaneously improving output quality.

Credit to @ehartford @fernandofernandes @DavidGF for laserRMT

Resources:
☄️ AutoLaser: https://colab.research.google.com/drive/11j0e-w6BfvqeFN1gUrpOqdW0vcKqfVqP?usp=sharing
laserRMT: https://github.com/cognitivecomputations/laserRMT
The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction (2312.13558)
·
replied to macadeliccc's post 10 months ago