GroqFlow™ is the easiest way to get started with Groq's technology. GroqFlow provides an automated tool flow for compiling machine learning and linear algebra workloads into Groq programs and executing those programs on GroqChip™ processors.
We recommend that your system meets the following hardware requirements:
- To build models: 32GB or more of RAM.
- To run models: 8 GroqChip processors is recommended, especially for larger models.
Sign-up on support.groq.com to download and install GroqWare™ Suite version >=0.9.2.1.
For installation instructions, please have a look at our Install Guide.
To Groq a PyTorch model, simply provide your model and inputs to the groqit() function. Once groqit() has built your model, you can execute your Groq model the same way you execute a PyTorch model.
groqit() also works with ONNX files and provides many customization options. You can find more information about how to use groqit() in our User Guide.
-
demo_helpers: Scripts used for GroqFlow demos and proof points.
-
docs: All information you'd need to be successful with GroqFlow.
-
examples: Includes various GroqFlow examples.
-
groqflow: The source code for the
groqflowpackage. -
proof_points: Machine learning proof points using GroqFlow.
-
readme.md: This readme.
-
setup.py: GroqFlow setup script for installation.
GroqFlow development is primarily conducted within Groq's internal repo and is periodically synced to GitHub. This approach means that developer contributions are not immediately obvious in the commit log. The following awesome developers have contributed to GroqFlow (order is alphabetical):
 Daniel Holanda |
 Jeremy Fowers |
 Lev Zlotnik |
 Philip Colangelo |
 Ramakrishnan Sivakumar |
 Sarah Garrod Massengill |
 Victoria Godsoe |
This project follows the all-contributors specification. Contributions of any kind are welcome!