Fix NestedTensor deprecation warning in SDPA tutorial

dorbittonn · dorbittonn · commit 05d7dc276a0c · 2025-11-28T21:19:09.000+02:00
Add layout=torch.jagged parameter to nested_tensor call to suppress
the prototype API warning and use the recommended jagged layout.
diff --git a/intermediate_source/scaled_dot_product_attention_tutorial.py b/intermediate_source/scaled_dot_product_attention_tutorial.py
@@ -223,7 +223,7 @@ def generate_rand_batch(
                 torch.randn(seq_len, embed_dimension,
                             dtype=dtype, device=device)
                 for seq_len in seq_len_list
-            ]
+            ], layout=torch.jagged
         ),
         seq_len_list,
     )

Original file line number	Diff line number	Diff line change
`@@ -223,7 +223,7 @@ def generate_rand_batch(`
`223`	`223`	`torch.randn(seq_len, embed_dimension,`
`224`	`224`	`dtype=dtype, device=device)`
`225`	`225`	`for seq_len in seq_len_list`
`226`		`- ]`
	`226`	`+ ], layout=torch.jagged`
`227`	`227`	`),`
`228`	`228`	`seq_len_list,`
`229`	`229`	`)`