> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cellulose.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

At time of this writing, most modern ML inference workloads,
especially ones in production, run on NVIDIA GPUs for better performance.
NVIDIA *TensorRT* is one such tool that takes in a
[PyTorch](https://github.com/pytorch/TensorRT) or
[ONNX](https://github.com/onnx/onnx-tensorrt) model, applies some model
optimizations and generate TensorRT binaries to be run on compatible GPUs.

Some model optimization techniques here include:

1. Weight and activation precision calibrations (quantization)
2. Operation (op) / layer fusions
3. Auto-tuning of kernels so only the best algorithms are selected to run on
   your specific target device (GPU)
4. Multi stream execution etc.

Cellulose currently integrates basic *TensorRT* features within our dashboard.
These include:

<CardGroup cols={2}>
  <Card title="Operator Annotations" icon="pen" href="https://docs.cellulose.ai/runtime-support/tensorrt/annotations">
    Figure out *TensorRT* compatibility on a per operation / layer basis
    at a glance from the model visualizer.
  </Card>

  <Card title="Quantization Workflows (coming soon)" icon="bars-filter" href="https://docs.cellulose.ai/runtime-support/tensorrt/quantization">
    Quantize your model with *TensorRT*, understand your accuracy vs. system
    performance improvement tradeoffs with ease.
  </Card>

  <Card title="Operator Fusion (coming soon)" icon="merge" href="https://docs.cellulose.ai/runtime-support/tensorrt/op-fusion">
    Understand which operator / layer fusions are automatically applied by
    *TensorRT*.
  </Card>
</CardGroup>

Read more about NVIDIA *TensorRT* [here](https://developer.nvidia.com/tensorrt)

## Have questions / need help?

Please reach out to [support@cellulose.ai](mailto:support@cellulose.ai), and we'll get back to you as soon
as possible.