DeepGEMM
PulseAugur coverage of DeepGEMM — every cluster mentioning DeepGEMM across labs, papers, and developer communities, ranked by signal.
- developed by DeepInfer: Open-Source Deep Learning Deployment Toolkit for Image-Guided Therapy 95%
- developed FlashMLA 90%
- uses DeepSeek-V4.1-Flash 90%
- developed by DeepSeek-V4.1-Flash 90%
- developed Seph Martin 90%
- competes with Seph Martin 70%
- competes with FlashMLA 70%
- used by FlashMLA 70%
- instance of DeepInfer: Open-Source Deep Learning Deployment Toolkit for Image-Guided Therapy 70%
- used by Seph Martin 70%
- affiliated with Seph Martin 70%
- used by DeepSeek-V4.1-Flash 70%
5 day(s) with sentiment data
-
TensorLean framework verifies GPU kernels, finds 38 bugs
Researchers from Columbia University and CertiK have developed TensorLean, a formal verification framework designed to ensure the correctness of GPU kernels at the PTX level. This Lean-based system analyzes compiled ker…
-
New NVLink technique speeds up Mixture-of-Experts AI inference
Researchers have developed NVLinkOneSided, a novel approach for accelerating inference in Mixture-of-Experts (MoE) models over NVLink. This method optimizes the expert parallelism all-to-all communication by enabling di…
-
New NVLinkOneSided protocol accelerates MoE inference on GPUs
Researchers have developed NVLinkOneSided, a novel one-sided expert-parallel all-to-all communication protocol designed for Mixture-of-Experts (MoE) inference on GPUs within a single NVLink domain. This method allows GP…
-
vLLM 0.33.0 ships stable ABI CUDA wheel for easier PyTorch updates
The vLLM team has released version 0.33.0, featuring a CUDA wheel built with PyTorch's stable ABI. This change aims to allow users to update their PyTorch versions without needing to recompile vLLM. The audit confirmed …
-
Prime Intellect boosts MoE training speed with fused backward pass kernel
Prime Intellect has enhanced DeepSeek's Mega MoE kernel by adding a fused, overlapping backward pass, making it suitable for training Mixture-of-Experts (MoE) models. This new kernel, Prime Mega MoE, significantly speed…
-
DeepSeek releases DeepInfer inference engine for optimized large-scale AI deployment
DeepSeek has open-sourced DeepInfer, a new inference engine designed to optimize large-scale model deployment. The engine incorporates advanced techniques such as a Zero-Overhead Batch Scheduler, asynchronous batching, …
-
DeepInfer vs. vLLM: Benchmarking inference speed on B200 GPUs
A new open-source inference framework called DeepInfer, developed by DeepSeek, has been benchmarked against vLLM on eight NVIDIA B200 GPUs. The integration, facilitated by GPUStack and a lightweight serving project call…
-
DeepSeek releases open-source DeepInfer inference framework
DeepSeek has released DeepInfer, an open-source inference framework under the Apache 2.0 license. This framework is designed for large-scale inference and powers DeepSeek's production services, with the open-source vers…
-
vLLM v0.31.0 enhances serving with FlashMLA, DeepGEMM, and MXFP8
vLLM has released version 0.31.0, bringing significant enhancements to its serving capabilities. This update integrates FlashMLA with V4.1 NVFP4 compressed KV cache as the default for SM100, alongside the inclusion of D…
-
DeepSeek open-sources AI infrastructure for Huawei Ascend hardware
DeepSeek has released open-source Ascend-compatible versions of its TileLang, DeepGEMM, and DeepEP infrastructure components. These tools are designed to optimize AI model performance on Huawei Ascend hardware. The rele…
-
DeepSeek, Huawei Open-Source Tools to Challenge Nvidia's AI Dominance · 10 sources tracked
DeepSeek and Huawei have jointly released a suite of open-source tools designed to enhance the software ecosystem for Huawei's Ascend AI chips. This initiative aims to provide a viable alternative to Nvidia's dominant C…
-
Zhipu AI's GLM-5.2 model deployed on serverless GPUs
Zhipu AI has released GLM-5.2, a 700B Mixture-of-Experts (MoE) model that excels in complex reasoning and software engineering tasks, reportedly matching or surpassing proprietary models like Claude 3.5 Sonnet and GPT-4…
-
DeepSeek V4 achieves faster performance with custom kernels, replacing cuBLAS
DeepSeek has developed a custom kernel stack, DeepGEMM and TileLang, which not only matches but surpasses the performance of NVIDIA's cuBLAS. This custom implementation achieves bitwise determinism and batch invariance,…