From 052d5e6d5f4a89e71ad84e3a9c81ad420597d914 Mon Sep 17 00:00:00 2001 From: Sayak Paul Date: Wed, 18 Mar 2026 15:43:53 +0530 Subject: [PATCH 1/2] Update attention_backends.md --- docs/source/en/optimization/attention_backends.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/source/en/optimization/attention_backends.md b/docs/source/en/optimization/attention_backends.md index f3ff4781c6ec..d239df4347ee 100644 --- a/docs/source/en/optimization/attention_backends.md +++ b/docs/source/en/optimization/attention_backends.md @@ -35,7 +35,7 @@ The [`~ModelMixin.set_attention_backend`] method iterates through all the module The example below demonstrates how to enable the `_flash_3_hub` implementation for FlashAttention-3 from the [`kernels`](https://github.com/huggingface/kernels) library, which allows you to instantly use optimized compute kernels from the Hub without requiring any setup. > [!NOTE] -> FlashAttention-3 is not supported for non-Hopper architectures, in which case, use FlashAttention with `set_attention_backend("flash")`. +> For FlashAttention-3, at least Ampere GPUs is needed. ```py import torch From 611034eb74153a1caf573000718abdba2dcd1d36 Mon Sep 17 00:00:00 2001 From: Sayak Paul Date: Wed, 18 Mar 2026 23:31:40 +0530 Subject: [PATCH 2/2] Update docs/source/en/optimization/attention_backends.md Co-authored-by: Steven Liu <59462357+stevhliu@users.noreply.github.com> --- docs/source/en/optimization/attention_backends.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/source/en/optimization/attention_backends.md b/docs/source/en/optimization/attention_backends.md index d239df4347ee..00be238383f9 100644 --- a/docs/source/en/optimization/attention_backends.md +++ b/docs/source/en/optimization/attention_backends.md @@ -35,7 +35,7 @@ The [`~ModelMixin.set_attention_backend`] method iterates through all the module The example below demonstrates how to enable the `_flash_3_hub` implementation for FlashAttention-3 from the [`kernels`](https://github.com/huggingface/kernels) library, which allows you to instantly use optimized compute kernels from the Hub without requiring any setup. > [!NOTE] -> For FlashAttention-3, at least Ampere GPUs is needed. +> FlashAttention-3 requires Ampere GPUs at a minimum. ```py import torch