kernels/pr_496/en/api/kernels.md

# Kernels API Reference

## Main Functions

### get_kernel[[kernels.get_kernel]]

#### kernels.get_kernel[[kernels.get_kernel]]

[Source](https://github.com/huggingface/kernels/blob/vr_496/kernels/src/kernels/utils.py#L301)

Load a kernel from the kernel hub.

This function downloads a kernel to the local Hugging Face Hub cache directory (if it was not downloaded before)
and then loads the kernel.

Example:
```python
import torch
from kernels import get_kernel

activation = get_kernel("kernels-community/relu", version=1)
x = torch.randn(10, 20, device="cuda")
out = torch.empty_like(x)
result = activation.relu(out, x)
```

**Parameters:**

repo_id (`str`) : The Hub repository containing the kernel.

revision (`str`, *optional*, defaults to `"main"`) : The specific revision (branch, tag, or commit) to download. Cannot be used together with `version`.

version (`int`, *optional*) : The kernel version to download. Cannot be used together with `revision`.

backend (`str`, *optional*) : The backend to load the kernel for. Can only be `cpu` or the backend that Torch is compiled for. The backend will be detected automatically if not provided.

user_agent (`Union[str, dict]`, *optional*) : The `user_agent` info to pass to `snapshot_download()` for internal telemetry.

**Returns:**

``ModuleType``

The imported kernel module.

### get_local_kernel[[kernels.get_local_kernel]]

#### kernels.get_local_kernel[[kernels.get_local_kernel]]

[Source](https://github.com/huggingface/kernels/blob/vr_496/kernels/src/kernels/utils.py#L360)

Import a kernel from a local kernel repository path.

**Parameters:**

repo_path (`Path`) : The local path to the kernel repository.

package_name (`str`) : The name of the package to import from the repository.

backend (`str`, *optional*) : The backend to load the kernel for. Can only be `cpu` or the backend that Torch is compiled for. The backend will be detected automatically if not provided.

**Returns:**

``ModuleType``

The imported kernel module.

### has_kernel[[kernels.has_kernel]]

#### kernels.has_kernel[[kernels.has_kernel]]

[Source](https://github.com/huggingface/kernels/blob/vr_496/kernels/src/kernels/utils.py#L396)

Check whether a kernel build exists for the current environment (Torch version and compute framework).

**Parameters:**

repo_id (`str`) : The Hub repository containing the kernel.

revision (`str`, *optional*, defaults to `"main"`) : The specific revision (branch, tag, or commit) to download. Cannot be used together with `version`.

version (`int`, *optional*) : The kernel version to download. Cannot be used together with `revision`.

backend (`str`, *optional*) : The backend to load the kernel for. Can only be `cpu` or the backend that Torch is compiled for. The backend will be detected automatically if not provided.

**Returns:**

``bool``

`True` if a kernel is available for the current environment.

### get_loaded_kernels[[kernels.get_loaded_kernels]]

#### kernels.get_loaded_kernels[[kernels.get_loaded_kernels]]

[Source](https://github.com/huggingface/kernels/blob/vr_496/kernels/src/kernels/utils.py#L54)

Return a snapshot of every kernel that has been loaded into the current process.

Each entry is a `kernels.utils.LoadedKernel` dataclass with fields:

- `kernel_id` (`str`): unique identifier used as the `sys.modules` key
  for this variant (either `metadata.id` or a hash-suffixed module name).
- `module` (`ModuleType`): the imported kernel module.
- `module_name` (`str`): the kernel's module name.
- `repo_infos` (`kernels.utils.RepoInfos | None`): populated only for
  kernels loaded via `get_kernel`. Loaders that work from a local path
  (`get_local_kernel`) or a lockfile (`get_locked_kernel`, `load_kernel`)
  leave this as `None`.

`RepoInfos` has `repo_id`, `revision`, and `backend` fields. `backend`
reflects the value passed by the caller — it is `None` when the caller
relied on backend auto-detection.

The returned list is a new list; mutating it does not affect the registry.

> [!NOTE]
> These arguments might be renamed / changed a bit.

Example:
```python
from kernels import get_kernel, get_loaded_kernels

get_kernel("kernels-community/activation", version=1)
for loaded in get_loaded_kernels():
    print(loaded.module_name, loaded.repo_infos)
```

**Returns:**

``list[LoadedKernel]``

one entry per distinct kernel variant path
loaded in this process.

## Loading locked kernels

### load_kernel[[kernels.load_kernel]]

#### kernels.load_kernel[[kernels.load_kernel]]

[Source](https://github.com/huggingface/kernels/blob/vr_496/kernels/src/kernels/utils.py#L441)

Get a pre-downloaded, locked kernel.

If `lockfile` is not specified, the lockfile will be loaded from the caller's package metadata.

**Parameters:**

repo_id (`str`) : The Hub repository containing the kernel.

lockfile (`Path`, *optional*) : Path to the lockfile. If not provided, the lockfile will be loaded from the caller's package metadata.

backend (`str`, *optional*) : The backend to load the kernel for. Can only be `cpu` or the backend that Torch is compiled for. The backend will be detected automatically if not provided.

**Returns:**

``ModuleType``

The imported kernel module.

### get_locked_kernel[[kernels.get_locked_kernel]]

#### kernels.get_locked_kernel[[kernels.get_locked_kernel]]

[Source](https://github.com/huggingface/kernels/blob/vr_496/kernels/src/kernels/utils.py#L514)

Get a kernel using a lock file.

**Parameters:**

repo_id (`str`) : The Hub repository containing the kernel.

local_files_only (`bool`, *optional*, defaults to `False`) : Whether to only use local files and not download from the Hub.

**Returns:**

``ModuleType``

The imported kernel module.