Large language models (LLMs) have demonstrated impressive abilities in various domains while the inference cost is expensive. The state-of-the-art methods use 2-bit quantization for mainstream LLMs. However, challenges still exist: (1) Nonnegligible accuracy loss for 2-bit quantization. Weights are quantized by groups, while the ranges of weights are large in some groups, resulting in large quantization errors and nonnegligible accuracy loss (e.g. >3% for Llama2-7b with 2-bit quantization in GPTQ and Greenbit). (2) Limited accuracy improvement by adding 4-bit weights. Increasing 10% extra average bit more 4-bit weights only leads to <0.5% accuracy improvement on a quantized Llama2-7b. (3) Time-consuming dequantization operations on GPUs. The dequantization operations lead to >50% execution time, hindering the potential of reducing LLM inference cost. To tackle these challenges, we propose the following techniques: (1) We only quantize a small fraction of groups with the larger range using 4-bit with memory alignment consideration on GPUs.(2) We design the asynchronous dequantization on GPUs, leading to up to 3.92X speedup. We conduct extensive experiments on different model sizes. We achieve 2.85-bit for each weight and the end-to-end speedup for Llama2-7b is 1.74X over the original model, and we reduce both runtime cost and hardware cost by up to 2.70X and 2.81X with less GPU requirements.

Extending image-based Large Multimodal Models (LMMs) to videos is challenging due to the inherent complexity of video data. The recent approaches extending image-based LMMs to videos either lack the grounding capabilities (e.g., VideoChat, Video-ChatGPT, Video-LLaMA) or do not utilize the audio-signals for better video understanding (e.g., Video-ChatGPT). Addressing these gaps, we propose PG-Video-LLaVA, the first LMM with pixel-level grounding capability, integrating audio cues by transcribing them into text to enrich video-context understanding. Our framework uses an off-the-shelf tracker and a novel grounding module, enabling it to spatially localize objects in videos following user instructions. We evaluate PG-Video-LLaVA using video-based generative and question-answering benchmarks and introduce new benchmarks specifically designed to measure prompt-based object grounding performance in videos. Further, we propose the use of Vicuna over GPT-3.5, as utilized in Video-ChatGPT, for video-based conversation benchmarking, ensuring reproducibility of results which is a concern with the proprietary nature of GPT-3.5. Our framework builds on SoTA image-based LLaVA model and extends its advantages to the video domain, delivering promising gains on video-based conversation and grounding tasks. Project Page: https://github.com/mbzuai-oryx/Video-LLaVA

Ensuring the safety of artificial intelligence-generated content (AIGC) is a longstanding topic in the artificial intelligence (AI) community, and the safety concerns associated with Large Language Models (LLMs) have been widely investigated. Recently, large vision-language models (VLMs) represent an unprecedented revolution, as they are built upon LLMs but can incorporate additional modalities (e.g., images). However, the safety of VLMs lacks systematic evaluation, and there may be an overconfidence in the safety guarantees provided by their underlying LLMs. In this paper, to demonstrate that introducing additional modality modules leads to unforeseen AI safety issues, we propose FigStep, a straightforward yet effective jailbreaking algorithm against VLMs. Instead of feeding textual harmful instructions directly, FigStep converts the harmful content into images through typography to bypass the safety alignment within the textual module of the VLMs, inducing VLMs to output unsafe responses that violate common AI safety policies. In our evaluation, we manually review 46,500 model responses generated by 3 families of the promising open-source VLMs, i.e., LLaVA, MiniGPT4, and CogVLM (a total of 6 VLMs). The experimental results show that FigStep can achieve an average attack success rate of 82.50% on 500 harmful queries in 10 topics. Moreover, we demonstrate that the methodology of FigStep can even jailbreak GPT-4V, which already leverages an OCR detector to filter harmful queries. Above all, our work reveals that VLMs are vulnerable to jailbreaking attacks, which highlights the necessity of novel safety alignments between visual and textual modalities.

We present a novel method, the Chain of Empathy (CoE) prompting, that utilizes insights from psychotherapy to induce Large Language Models (LLMs) to reason about human emotional states. This method is inspired by various psychotherapy approaches including Cognitive Behavioral Therapy (CBT), Dialectical Behavior Therapy (DBT), Person Centered Therapy (PCT), and Reality Therapy (RT), each leading to different patterns of interpreting clients' mental states. LLMs without reasoning generated predominantly exploratory responses. However, when LLMs used CoE reasoning, we found a more comprehensive range of empathetic responses aligned with the different reasoning patterns of each psychotherapy model. The CBT based CoE resulted in the most balanced generation of empathetic responses. The findings underscore the importance of understanding the emotional context and how it affects human and AI communication. Our research contributes to understanding how psychotherapeutic models can be incorporated into LLMs, facilitating the development of context-specific, safer, and empathetic AI.

Massive transformer-based models face several challenges, including slow and computationally intensive pre-training and over-parametrization. This paper addresses these challenges by proposing a versatile method called GQKVA, which generalizes query, key, and value grouping techniques. GQKVA is designed to speed up transformer pre-training while reducing the model size. Our experiments with various GQKVA variants highlight a clear trade-off between performance and model size, allowing for customized choices based on resource and time limitations. Our findings also indicate that the conventional multi-head attention approach is not always the best choice, as there are lighter and faster alternatives available. We tested our method on ViT, which achieved an approximate 0.3% increase in accuracy while reducing the model size by about 4% in the task of image classification. Additionally, our most aggressive model reduction experiment resulted in a reduction of approximately 15% in model size, with only around a 1% drop in accuracy.

The capabilities of large language models (LLMs) have sparked debate over whether such systems just learn an enormous collection of superficial statistics or a coherent model of the data generation process -- a world model. We find preliminary evidence for the latter by analyzing the learned representations of three spatial datasets (world, US, NYC places) and three temporal datasets (historical figures, artworks, news headlines) in the Llama-2 family of models. We discover that LLMs learn linear representations of space and time across multiple scales. These representations are robust to prompting variations and unified across different entity types (e.g. cities and landmarks). In addition, we identify individual ``space neurons'' and ``time neurons'' that reliably encode spatial and temporal coordinates. While further investigation is needed, our results suggest modern LLMs learn rich spatiotemporal representations of the real world and possess basic ingredients of a world model.

As the size of large language models (LLMs) continues to grow, model compression without sacrificing accuracy has become a crucial challenge for deployment. While some quantization methods, such as GPTQ, have made progress in achieving acceptable 4-bit weight-only quantization, attempts at lower-bit quantization often result in severe performance degradation. In this paper, we introduce a technique called norm tweaking, which can be used as a plugin in current PTQ methods to achieve high precision while being cost-efficient. Our approach is inspired by the observation that rectifying the quantized activation distribution to match its float counterpart can readily restore accuracy for LLMs. To achieve this, we carefully design a tweaking strategy that includes calibration data generation and channel-wise distance constraint to update the weights of normalization layers for better generalization. We conduct extensive experiments on various datasets using several open-sourced LLMs. Our method demonstrates significant improvements in both weight-only quantization and joint quantization of weights and activations, surpassing existing PTQ methods. On GLM-130B and OPT-66B, our method even achieves the same level of accuracy at 2-bit quantization as their float ones. Our simple and effective approach makes it more practical for real-world applications.

Large language models such as GPT-3 & ChatGPT have transformed deep learning (DL), powering applications that have captured the public's imagination. These models are rapidly being adopted across domains for analytics on various modalities, often by finetuning pre-trained base models. Such models need multiple GPUs due to both their size and computational load, driving the development of a bevy of "model parallelism" techniques & tools. Navigating such parallelism choices, however, is a new burden for end users of DL such as data scientists, domain scientists, etc. who may lack the necessary systems knowhow. The need for model selection, which leads to many models to train due to hyper-parameter tuning or layer-wise finetuning, compounds the situation with two more burdens: resource apportioning and scheduling. In this work, we tackle these three burdens for DL users in a unified manner by formalizing them as a joint problem that we call SPASE: Select a Parallelism, Allocate resources, and SchedulE. We propose a new information system architecture to tackle the SPASE problem holistically, representing a key step toward enabling wider adoption of large DL models. We devise an extensible template for existing parallelism schemes and combine it with an automated empirical profiler for runtime estimation. We then formulate SPASE as an MILP. We find that direct use of an MILP-solver is significantly more effective than several baseline heuristics. We optimize the system runtime further with an introspective scheduling approach. We implement all these techniques into a new data system we call Saturn. Experiments with benchmark DL workloads show that Saturn achieves 39-49% lower model selection runtimes than typical current DL practice.

The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15.5B parameter models with 8K context length, infilling capabilities and fast large-batch inference enabled by multi-query attention. StarCoderBase is trained on 1 trillion tokens sourced from The Stack, a large collection of permissively licensed GitHub repositories with inspection tools and an opt-out process. We fine-tuned StarCoderBase on 35B Python tokens, resulting in the creation of StarCoder. We perform the most comprehensive evaluation of Code LLMs to date and show that StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40\% pass@1 on HumanEval, and still retains its performance on other programming languages. We take several important steps towards a safe open-access model release, including an improved PII redaction pipeline and a novel attribution tracing tool, and make the StarCoder models publicly available under a more commercially viable version of the Open Responsible AI Model license.

Knowledge-based visual question answering (VQA) requires external knowledge beyond the image to answer the question. Early studies retrieve required knowledge from explicit knowledge bases (KBs), which often introduces irrelevant information to the question, hence restricting the performance of their models. Recent works have resorted to using a powerful large language model (LLM) as an implicit knowledge engine to acquire the necessary knowledge for answering. Despite the encouraging results achieved by these methods, we argue that they have not fully activated the capacity of the blind LLM as the provided textual input is insufficient to depict the required visual information to answer the question. In this paper, we present Prophet -- a conceptually simple, flexible, and general framework designed to prompt LLM with answer heuristics for knowledge-based VQA. Specifically, we first train a vanilla VQA model on a specific knowledge-based VQA dataset without external knowledge. After that, we extract two types of complementary answer heuristics from the VQA model: answer candidates and answer-aware examples. Finally, the two types of answer heuristics are jointly encoded into a formatted prompt to facilitate the LLM's understanding of both the image and question, thus generating a more accurate answer. By incorporating the state-of-the-art LLM GPT-3, Prophet significantly outperforms existing state-of-the-art methods on four challenging knowledge-based VQA datasets. To demonstrate the generality of our approach, we instantiate Prophet with the combinations of different VQA models (i.e., both discriminative and generative ones) and different LLMs (i.e., both commercial and open-source ones).