Speculative Decoding in vLLM

Speculative decoding in vLLM is a powerful technique that accelerates token generation by leveraging both small and large models in tandem. In this blog, we’ll break down speculative decoding in vLLM, how it works, and the performance improvements it brings. This content is based on a session from our bi-weekly vLLM Office Hours, where we discuss techniques and updates to optimize vLLM performance.

vLLM v0.6

vLLM achieves 2.7x higher throughput and 5x faster TPOT (time per output token) on Llama 8B model, and 1.8x higher throughput and 2x less TPOT on Llama 70B model. A month ago, we released our performance roadmap committing to performance as our top priority. We will start by diagnosing the performance bottleneck in vLLM previously.

SGLang v0.3

We’re excited to announce the release of SGLang v0.3, which brings significant performance enhancements and expanded support for novel model architectures. Here are the key updates: Up to 7x higher throughput for DeepSeek Multi-head Latent Attention (MLA). Up to 1.5x lower latency with torch.compile on small batch sizes. Support for interleaved text and multi-image/video in LLaVA-OneVision.

Qwen2-VL

Qwen2-VL is the latest version of the vision language models in the Qwen model families. SoTA understanding of images of various resolution & ratio: Qwen2-VL achieves state-of-the-art performance on visual understanding benchmarks, including MathVista, DocVQA, RealWorldQA, MTVQA, etc. Understanding videos of 20min+: with the online streaming capabilities, Qwen2-VL can understand videos over 20 minutes for high-quality video-based question answering, dialog, content creation, etc.

Text2SQL Leaderboard

Text-to-SQL (or Text2SQL), as the name implies, is to convert text into SQL. A more academic definition is to convert natural language problems in the database field into structured query languages ​​that can be executed in relational databases. Therefore, Text-to-SQL can also be abbreviated as NL2SQL. Input: natural language questions, such as Query the relevant information of the table t_user, and the results are sorted in descending order by id.

SGLang v0.2

Through our operational experiences and in-depth research, we’ve continuously enhanced the underlying serving systems for the Chatbot Arena platform, spanning from the high-level multi-model serving framework, FastChat, to the efficient serving engine, SGLang Runtime (SRT). This post focuses on SGLang Runtime, a general-purpose serving engine for LLMs and VLMs. While existing options like TensorRT-LLM, vLLM, MLC-LLM, and Hugging Face TGI have their merits, we found them sometimes hard to use, difficult to customize, or lacking in performance.

DeepSeek-Coder-V2

We present DeepSeek-Coder-V2, an open-source Mixture-of-Experts (MoE) code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks. Specifically, DeepSeek-Coder-V2 is further pre-trained from an intermediate checkpoint of DeepSeek-V2 with additional 6 trillion tokens. Through this continued pre-training, DeepSeek-Coder-V2 substantially enhances the coding and mathematical reasoning capabilities of DeepSeek-V2, while maintaining comparable performance in general language tasks.

DeepSeek-V2

We introduce DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token. Compared with DeepSeek 67B, DeepSeek-V2 achieves stronger performance, and meanwhile saves 42.5% of training costs, reduces the KV cache by 93.3%, and boosts the maximum generation throughput to more than 5 times.