helion.language.load

helion.language.load(tensor, index, extra_mask=None, eviction_policy=None)

Load a value from a tensor using a list of indices.

This function is equivalent to tensor[index] but allows setting extra_mask= to mask elements beyond the default masking based on the hl.tile range. It also accepts an optional eviction_policy which is forwarded to the underlying Triton tl.load call to control the cache eviction behavior (e.g., “evict_last”).

Parameters:

• tensor (Tensor | StackTensor) – The tensor / stack tensor to load from

• index (list[object]) – The indices to use to index into the tensor

• extra_mask (Tensor | None) – The extra mask (beyond automatic tile bounds masking) to apply to the tensor

• eviction_policy (str | None) – Optional Triton load eviction policy to hint cache behavior

Returns:

The loaded value

Return type:

torch.Tensor