Hamdi Amroun, PhD

Exploring Parameter-Efficient Fine-Tuning Strategies for Large Language Models Quick read: https://lnkd.in/gWKCv38m Researchers from Northeastern University, the University of California, Arizona State University, and New York University present this survey thoroughly examining diverse PEFT algorithms and evaluating their performance and computational requirements. It also provides an overview of applications developed using various PEFT methods and discusses common strategies employed to reduce computational expenses associated with PEFT. Beyond algorithmic considerations, the survey delves into real-world system designs to explore the implementation costs of different PEFT algorithms. As an invaluable resource, this survey equips researchers with insights into PEFT algorithms and their system implementations, offering detailed analyses of recent progressions and practical uses. The researchers categorized PEFT algorithms into additive, selective, reparameterized, and hybrid fine-tuning based on their operations. Major additive fine-tuning algorithms include adapters, soft prompts, and others, which differ in the additional tunable modules or parameters they utilize. Selective fine-tuning, in contrast, involves selecting a small subset of parameters from the backbone model, making only these parameters tunable while leaving the majority untouched during downstream task fine-tuning. Selective fine-tuning is categorized based on the grouping of chosen parameters: Unstructural Masking and Structural Masking. Reparametrization involves transforming model parameters between two equivalent forms, introducing additional low-rank trainable parameters during training, which are then integrated with the original model for inference. This approach encompasses two main strategies: Low-rank Decomposition and LoRA Derivatives. Hybrid fine-tuning explores different PEFT methods’ design spaces and combines their advantages. Paper: https://lnkd.in/gh8qfi4k

26

#EarlyCareer #talk at #IJCAI2024 by🗣️Julia Neidhart, Technische Universität Wien: Transforming Recommender Systems: Balancing Personalization, Fairness, and Human Values ➡️https://lnkd.in/dK3Nieyr Abstract: Recent advancements in recommender systems highlight metrics beyond accuracy, including diversity, serendipity and fairness. This paper examines the integration of Large Language Models into recommender systems, focusing on challenges such as interpretability, bias, and user autonomy. LLMs offer enhanced personalization capabilities but also introduce concerns regarding transparency and ethical implications. This work discusses ongoing research efforts aiming to develop transparent, fair, and contextually aware systems. Our approach seeks to prioritize user wellbeing and inclusivity, contributing to a more equitable and functional digital environment through advanced technologies and interdisciplinary insights. #XAI #LLMs #InclusiveTech #NextGenAI #AI TU Wien Informatics

3

Happy to share our latest paper, "𝐅𝐢𝐧𝐝𝐢𝐧𝐠 𝐚 𝐍𝐞𝐞𝐝𝐥𝐞 𝐢𝐧 𝐚 𝐇𝐚𝐲𝐬𝐭𝐚𝐜𝐤: A Machine Learning Framework for Anomaly Detection in Payment Systems", co-authored by Anneke Kosse and Jacob Sharples, now available as both the Bank of Canada (https://lnkd.in/e7ch4zeW) and the Bank for International Settlements – BIS (https://lnkd.in/eNdGE2iP) working paper. It employs a layered ML framework designed to effectively handle large volumes and address the scarcity of pre-labeled examples of anomalous transactions, providing new tools for safeguarding financial systems. #AI #machinelearning #anomalydetection #paymentsystems #transactionmonitoring

81

7 Comments

In the ever-evolving field of AI, efficiently finetuning large language models (LLMs) is crucial. Recent advancements, such as LLaMA-Adapters, LoRA (Low-Rank Adaptation), and its variant QLoRA, demonstrate significant strides in parameter-efficient finetuning. These methods reduce computational costs and make it possible for many of us to fine-tune these models. I came across three articles by Sebastian Raschka, PhD that I found very helpful. Article 1: Understanding LLaMA Adapters (https://lnkd.in/gFBdpH6n): Explores parameter-efficient finetuning methods for large language models (LLMs), highlighting prefix tuning, adapters, and the LLaMA-Adapter. Article 2: Parameter-Efficient Finetuning with LoRA (https://lnkd.in/gDvGbYkc): Discusses LoRA (Low-Rank Adaptation), which enables efficient finetuning of LLMs by reducing the number of trainable parameters. This method helps maintain performance while reducing computational costs and energy consumption. Article 3: Insights on LoRA and QLoRA (https://lnkd.in/gPzRWenM): Provides insights from extensive experiments on LoRA and its variant QLoRA. Highlights the effectiveness of these methods in various scenarios and their potential for broader applications in AI model adaptation.

5

👨‍💻 👩‍💻 Empowering data science teams with collaboration & efficiency! #Nebari, an open source platform, is making waves. We spoke with Kim Pevey from Quansight to get the inside scoop. What can Nebari do for you? Collaboration tools, environment management, extensions & more! 👀 Read the full blog & tell us what you think! Is Nebari a good fit for your team? https://lnkd.in/eAUcz92E #DataScience #OpenSource #Quansight

14

Explore the Exciting World of Deep Learning with ProjectPro's Step-by-Step Project-Based Deep Learning Roadmap that covers projects spanning foundational principles of neural networks to advanced techniques like Attention Mechanisms and Generative Models - https://bit.ly/3W4kWd2 #deeplearning #learndeeplearning #genai #generativemodels 

6

Researchers at the University of Waterloo Introduce Orchid: Revolutionizing Deep Learning with Data-Dependent Convolutions for Scalable Sequence Modeling In deep learning, especially in NLP, image analysis, and biology, there is an increasing focus on developing models that offer both computational efficiency and robust expressiveness. Attention mechanisms have been revolutionary, allowing for better... https://lnkd.in/eavf-sCk #AI #ML #Automation

1

LLAMA-NAS: Efficient Neural Architecture Search - Optimized LLMs: LLAMA-NAS proposes a method for creating efficient, smaller LLMs using one-shot NAS. - Significant Compression: Achieves 1.5x reduction in model size and 1.3x speedup with minimal accuracy loss. - Complementary Techniques: Method works well with quantization, further reducing model size and complexity. Subscribe to our daily newsletter here for more AI news https://lnkd.in/d-p9DiUr #Innovation #Technology #Future #VentureCapital #Entrepreneurship #AI #Data #MachineLearning #DeepLearning #NeuralNetworks

1

2 Comments

Peer into the exciting world of Machine Learning (ML) and Deep Learning (DL) research! This poster showcases 10 cutting-edge areas where researchers are pushing the boundaries of what's possible!!

3

There is a lot of conversation on X about how llms just aren't intelligent enough. A new paper, "Can language models be used as world simulators?" on Arxiv. The paper asks the question: can current llms accurately serve as world simulators by accurately predicting how actions change world states? it findings are that we are some distance away from skynet type scenarios. current got 4 for example is only ~60% accurate at simulating state changes based on common-sense tasks, like boiling water. #genai #ai

1

1 Comment

Can LLMs Teach Themselves? Exploring Self-Improvement in Language Model Training If so, which generation of LLMs can do so? The current generation IME cannot. https://lnkd.in/gK-cevmy

4

Exciting to see that folks are speaking our language! Thanks for the mention Neo4j 🚀 Check out our recent benchmark study cited in the blog post: https://lnkd.in/epJMkBQa

6

Google DeepMind Introduces Zipper: A Multi-Tower Decoder Architecture for Fusing Modalities Integrating multiple generative foundation models helps by combining the strengths of models trained on different modalities, such as text, speech, and images, enabling the system to perform cross-modal tasks effectively. This integration allows for... https://lnkd.in/eEAM_dKj #AI #ML #Automation

1

From UC Berkeley, Stanford, Carnegie Mellon, & Google Deepmind Octo An Open-Source Generalist Robot Policy https://lnkd.in/ePvtAAeg Large policies pretrained on diverse robot datasets have the potential to transform robotic learning: instead of training new policies from scratch, such generalist robot policies may be finetuned with only a little in-domain data, yet generalize broadly. However, to be widely applicable across a range of robotic learning scenarios, environments, and tasks, such policies need to handle diverse sensors and action spaces, accommodate a variety of commonly used robotic platforms, and finetune readily and efficiently to new domains. In this work, we aim to lay the groundwork for developing open-source, widely applicable, generalist policies for robotic manipulation. As a first step, we introduce Octo, a large transformer-based policy trained on 800k trajectories from the Open X-Embodiment dataset, the largest robot manipulation dataset to date. It can be instructed via language commands or goal images and can be effectively finetuned to robot setups with new sensory inputs and action spaces within a few hours on standard consumer GPUs. In experiments across 9 robotic platforms, we demonstrate that Octo serves as a versatile policy initialization that can be effectively finetuned to new observation and action spaces. We also perform detailed ablations of design decisions for the Octo model, from architecture to training data, to guide future research on building generalist robot models. #AI #robots #opensource #HuggingFace

15