misc: fix kv-layout doc references (#1009)

Author: Edenzzzz

flashinfer/cascade.py

@@ -568,7 +568,7 @@ class BatchDecodeWithSharedPrefixPagedKVCacheWrapper:
     of requests. The shared-prefix KV-Cache was stored in a standalone tensors, and the
     unique KV-Cache of each request was stored in a paged KV-Cache data structure.
 
-    Check :ref:`our tutorial<page-layout>` for page table layout.
+    Check :ref:`our tutorial<kv-layout>` for page table layout.
 
     Warning
     -------
@@ -807,7 +807,7 @@ class BatchPrefillWithSharedPrefixPagedKVCacheWrapper:
     r"""Wrapper class for prefill/append attention with shared-prefix paged kv-cache for
     batch of requests.
 
-    Check :ref:`our tutorial<page-layout>` for paged kv-cache layout.
+    Check :ref:`our tutorial<kv-layout>` for paged kv-cache layout.
 
     Warning
     -------

flashinfer/decode.py

@@ -509,7 +509,7 @@ class BatchDecodeWithPagedKVCacheWrapper:
     r"""Wrapper class for decode attention with paged kv-cache (first proposed in
     `vLLM <https://arxiv.org/abs/2309.06180>`_) for batch of requests.
 
-    Check :ref:`our tutorial<page-layout>` for page table layout.
+    Check :ref:`our tutorial<kv-layout>` for page table layout.
 
     Examples
     --------
@@ -1187,7 +1187,7 @@ class CUDAGraphBatchDecodeWithPagedKVCacheWrapper(BatchDecodeWithPagedKVCacheWra
     because we won't dispatch to different kernels for different batch sizes/sequence lengths/etc
     to accommodate the CUDAGraph requirement.
 
-    Check :ref:`our tutorial<page-layout>` for page table layout.
+    Check :ref:`our tutorial<kv-layout>` for page table layout.
 
     Note
     ----

flashinfer/gemm.py

@@ -449,7 +449,7 @@ def run(
             y[i] = x[i] \times W[\text{weight_indices}[i]]
 
         We use Ragged Tensor to represent the input tensor :attr:`x` and the output tensor :attr:`y`, and each x[i]
-        is a segment of the concatenated tensor. Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details.
+        is a segment of the concatenated tensor. Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details.
         We use a ``seg_len`` or ``seg_indptr`` tensor (either would work) to indicate the start and end of each segment,
         where the ``seg_indptr`` is the cumulative sum of the ``seg_lens`` tensor (with an additional 0 at the beginning):
 

flashinfer/pod.py

@@ -78,7 +78,7 @@ class PODWithPagedKVCacheWrapper:
     r"""Wrapper class for POD-Attention with paged kv-cache (first proposed in
     `<https://arxiv.org/abs/2410.18038>`_) for batch of requests.
 
-    Check :ref:`our tutorial<page-layout>` for page table layout.
+    Check :ref:`our tutorial<kv-layout>` for page table layout.
 
     Examples
     --------

flashinfer/prefill.py

@@ -887,7 +887,7 @@ class BatchPrefillWithPagedKVCacheWrapper:
     r"""Wrapper class for prefill/append attention with paged kv-cache for batch of
     requests.
 
-    Check :ref:`our tutorial <page-layout>` for page table layout.
+    Check :ref:`our tutorial <kv-layout>` for page table layout.
 
     Example
     -------
@@ -1722,7 +1722,7 @@ class BatchPrefillWithRaggedKVCacheWrapper:
     r"""Wrapper class for prefill/append attention with ragged (tensor) kv-cache for
     batch of requests.
 
-    Check :ref:`our tutorial <ragged-layout>` for ragged kv-cache layout.
+    Check :ref:`our tutorial <kv-layout>` for ragged kv-cache layout.
 
     Example
     -------

flashinfer/rope.py

@@ -298,7 +298,7 @@ def apply_rope_inplace(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters
@@ -384,7 +384,7 @@ def apply_rope_pos_ids_inplace(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters
@@ -446,7 +446,7 @@ def apply_llama31_rope_inplace(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters
@@ -553,7 +553,7 @@ def apply_llama31_rope_pos_ids_inplace(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters
@@ -629,7 +629,7 @@ def apply_rope(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters
@@ -738,7 +738,7 @@ def apply_rope_pos_ids(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters
@@ -810,7 +810,7 @@ def apply_llama31_rope(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters
@@ -931,7 +931,7 @@ def apply_llama31_rope_pos_ids(
     segment the query of the i-th segment is ``q[indptr[i]:indptr[i+1]]`` and the key of the
     i-th segment is ``k[indptr[i]:indptr[i+1]]``, the first element of :attr:`indptr` is always
     0 and the last element of :attr:`indptr` is the total number of queries/keys in the batch.
-    Please see :ref:`Ragged Tensor tutorial <ragged-layout>` for more details about the
+    Please see :ref:`Ragged Tensor tutorial <kv-layout>` for more details about the
     ragged tensor.
 
     Parameters