Instructions to use replicate/aiter-kernels with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Kernels
How to use replicate/aiter-kernels with Kernels:
# !pip install kernels from kernels import get_kernel # a version (or an explicit revision) is required; see the "Files and versions" tab for the available ones kernel = get_kernel("replicate/aiter-kernels", version=1) - Notebooks
- Google Colab
- Kaggle
Download build/torch-rocm/comms/__init__.py from replicate/aiter-kernels: direct link, hf CLI and curl.
- Browser
- Download file 962 Bytes
-
https://huggingface.co/replicate/aiter-kernels/resolve/main/build/torch-rocm/comms/__init__.py
- Command line
-
hf download hf://replicate/aiter-kernels/build/torch-rocm/comms/__init__.py
-
curl -L -o __init__.py https://huggingface.co/replicate/aiter-kernels/resolve/main/build/torch-rocm/comms/__init__.py
962 Bytes
| # SPDX-License-Identifier: MIT | |
| # Copyright (C) 2025, Advanced Micro Devices, Inc. All rights reserved. | |
| """ | |
| Triton-based communication primitives for AITER. | |
| This submodule contains communication operations implemented using Triton, | |
| including Iris-based GPU-initiated communication. | |
| If Iris is not available, importing this module will raise ImportError. | |
| """ | |
| # Import all Iris-based communication primitives | |
| # If Iris is not installed, this import will fail and the entire | |
| # aiter.ops.triton.comms module will be unavailable | |
| from .iris import IrisCommContext, calculate_heap_size | |
| from .reduce_scatter import reduce_scatter | |
| from .all_gather import all_gather | |
| from .fused import reduce_scatter_rmsnorm_quant_all_gather | |
| __all__ = [ | |
| "IrisCommContext", | |
| "calculate_heap_size", | |
| "reduce_scatter", | |
| "all_gather", | |
| "reduce_scatter_rmsnorm_quant_all_gather", | |
| "IRIS_COMM_AVAILABLE", | |
| ] | |
| # If we got here, Iris is available | |
| IRIS_COMM_AVAILABLE = True | |