typo: fix a bunch of typos. (#862)

Author: didier-durand

Diff for: flashinfer/activation.py

@@ -104,7 +104,7 @@ def silu_and_mul(input: torch.Tensor, out: torch.Tensor = None) -> torch.Tensor:
         Input tensor, shape (..., 2 * hidden_size).
 
     out: Optional[torch.Tensor]
-        The the output tensor, if specified, the kernel will update this tensor inplace.
+        The output tensor, if specified, the kernel will update this tensor inplace.
 
     Returns
     -------
@@ -139,7 +139,7 @@ def gelu_tanh_and_mul(input: torch.Tensor, out: torch.Tensor = None) -> torch.Te
         Input tensor, shape (..., 2 * hidden_size).
 
     out: Optional[torch.Tensor]
-        The the output tensor, if specified, the kernel will update this tensor inplace.
+        The output tensor, if specified, the kernel will update this tensor inplace.
 
     Returns
     -------
@@ -171,7 +171,7 @@ def gelu_and_mul(input: torch.Tensor, out: torch.Tensor = None) -> torch.Tensor:
         Input tensor, shape (..., 2 * hidden_size).
 
     out: Optional[torch.Tensor]
-        The the output tensor, if specified, the kernel will update this tensor inplace.
+        The output tensor, if specified, the kernel will update this tensor inplace.
 
     Returns
     -------

Diff for: flashinfer/norm.py

@@ -61,7 +61,7 @@ def rmsnorm(
     eps: float
         Epsilon for numerical stability.
     out: Optional[torch.Tensor]
-        The the output tensor, if specified, the kernel will update this tensor inplace.
+        The output tensor, if specified, the kernel will update this tensor inplace.
 
     Returns
     -------
@@ -144,7 +144,7 @@ def gemma_rmsnorm(
     eps: float
         Epsilon for numerical stability.
     out: Optional[torch.Tensor]
-        The the output tensor, if specified, the kernel will update this tensor inplace.
+        The output tensor, if specified, the kernel will update this tensor inplace.
 
     Returns
     -------

Diff for: flashinfer/page.py

@@ -180,9 +180,9 @@ def get_batch_indices_positions(
     Returns
     -------
     batch_indices: torch.Tensor
-        The batch indices of the each entry in the ragged tensor, shape: ``[nnz]``.
+        The batch indices of each entry in the ragged tensor, shape: ``[nnz]``.
     positions: torch.Tensor
-        The positions of the each entry in the ragged tensor, shape: ``[nnz]``.
+        The positions of each entry in the ragged tensor, shape: ``[nnz]``.
 
     Example
     -------
@@ -201,7 +201,7 @@ def get_batch_indices_positions(
     ----
     This function is similar to `CSR2COO <https://docs.nvidia.com/cuda/cusparse/#csr2coo>`_
     conversion in cuSPARSE library, with the difference that we are converting from a ragged
-    tensor (which don't require a column indices array) to a COO format.
+    tensor (which doesn't require a column indices array) to a COO format.
 
     See Also
     --------
@@ -349,7 +349,7 @@ def append_paged_kv_cache(
 
     Note
     ----
-    The function assumes that the space for appended k/v have already been allocated,
+    The function assumes that the space for appended k/v has already been allocated,
     which means :attr:`kv_indices`, :attr:`kv_indptr`, :attr:`kv_last_page_len` has
     incorporated appended k/v.
 

Diff for: flashinfer/prefill.py

@@ -991,7 +991,7 @@ class BatchPrefillWithPagedKVCacheWrapper:
     Note
     ----
     To accelerate computation, FlashInfer's batch prefill/append attention operators
-    creates some auxiliary data structures, these data structures can be reused across
+    create some auxiliary data structures, these data structures can be reused across
     multiple prefill/append attention calls (e.g. different Transformer layers). This
     wrapper class manages the lifecycle of these data structures.
     """
@@ -1815,7 +1815,7 @@ class BatchPrefillWithRaggedKVCacheWrapper:
     Note
     ----
     To accelerate computation, FlashInfer's batch prefill/append attention operators
-    creates some auxiliary data structures, these data structures can be reused across
+    create some auxiliary data structures, these data structures can be reused across
     multiple prefill/append attention calls (e.g. different Transformer layers). This
     wrapper class manages the lifecycle of these data structures.
     """