Search
Sort Score
Result 10 results
Languages All
Labels All
Results 11 - 20 of 304 for

gpu

(0.34 sec)

  1. TensorSplit|Documentation

    Proportion of the model placed on each Gpu in Aspose.LLM for .NET — balances load across unequal-VRAM Gpus....an array of floats, one per GPU, that controls the proportion...model placed on each GPU during multi-GPU split. Values are normalized...

    docs.aspose.com/llm/net/developer-reference/par...
    more..

  2. Acceleration|Documentation

    Hardware acceleration backends supported by Aspose.LLM for .NET, CUDA, HIP (ROCm), Metal, Vulkan, and CPU variants. How auto-detection works and how to force a specific backend....per-acceleration variants: one for each GPU backend, plus CPU variants at...Supported backends Backend Platforms GPU vendor Typical priority CUDA...

    docs.aspose.com/llm/net/developer-reference/acc...
    more..

  3. HIP / ROCm|Documentation

    Run Aspose.LLM for .NET on AMD Gpus with HIP (ROCm) acceleration — supported Gpus, driver requirements, and tuning notes....ROCm stack) is the AMD-specific GPU backend for Aspose.LLM for .NET...at AMD Instinct data-center GPUs and recent Radeon consumer cards...

    docs.aspose.com/llm/net/developer-reference/acc...
    more..

  4. OffloadKqv|Documentation

    Offload the KQV (attention) computation and KV cache to Gpu in Aspose.LLM for .NET — true by default on Gpu builds....cache itself live on GPU memory. On GPU-enabled builds, the default...typically true on GPU builds) Category KV cache / GPU Field on ContextParameters...

    docs.aspose.com/llm/net/developer-reference/par...
    more..

  5. Supported acceleration|Documentation

    Hardware acceleration backends supported by Aspose.LLM for .NET, CUDA, HIP, Metal, Vulkan, and CPU, with a platform × backend matrix....(drivers, GpuLayers , multi-GPU split), see the detail pages...Typical use CUDA ✅ ✅ ❌ NVIDIA GPUs. Highest throughput on modern...

    docs.aspose.com/llm/net/product-overview/suppor...
    more..

  6. UseGpu|Documentation

    Offload the vision projector to Gpu in Aspose.LLM for .NET — null defers to mtmd's auto-detection; explicit true/false forces a choice....runs the vision projector on GPU. Relevant only for vision presets;...default — typically true if a GPU is present) Category Multimodal...

    docs.aspose.com/llm/net/developer-reference/par...
    more..

  7. Model inference parameters|Documentation

    Configure how a model loads into memory in Aspose.LLM for .NET — Gpu layer offload, tensor split across Gpus, memory mapping, memory locking, tensor validation, and metadata overrides....offload to GPU, how to split across multiple GPUs, whether to...of model layers to offload to GPU. 0 = CPU only, 999 = full offload...

    docs.aspose.com/llm/net/developer-reference/par...
    more..

  8. Vulkan|Documentation

    Run Aspose.LLM for .NET with Vulkan acceleration — cross-platform, cross-vendor Gpu backend that works on NVIDIA, AMD, and Intel hardware without vendor-specific runtimes....Hide ] Vulkan is a cross-vendor GPU backend that works on most modern...modern discrete and integrated GPUs. It does not need CUDA or ROCm...

    docs.aspose.com/llm/net/developer-reference/acc...
    more..

  9. Aspose OCR slow on virtual machine - Free Suppo...

    Hello, My team noticed that the Aspose OCR engine is much slower on our virtual machine than on my desktop computer when using the Aspose OCR .NET library to extract text content from files. This raises a couple questio…...computer with a GPU compared to one without a GPU? Note that we...using Aspose.OCR (not Aspose.OCR-GPU). The specifications for both...

    forum.aspose.com/t/aspose-ocr-slow-on-virtual-m...
    more..

  10. OpOffload|Documentation

    Offload host tensor operations to device in Aspose.LLM for .NET — supplementary to GpuLayers; leave at default unless instructed....host-side tensor operations to the GPU device. This is supplementary...otherwise run on CPU even with GPU offload active. Quick reference...

    docs.aspose.com/llm/net/developer-reference/par...
    more..