Fix typos in comments and documentation (#4662)

Summary:
Pull Request resolved: #4662

as the title suggests

Reviewed By: alibeklfc

Differential Revision: D86405492

fbshipit-source-id: cfc7bb0c182a967ef2237e652a266743100c77e6

Author: Subhadeep Karan

faiss/gpu/GpuIndexBinaryCagra.h

@@ -111,7 +111,7 @@ struct GpuIndexBinaryCagra : public IndexBinary {
             const SearchParameters* search_params) const;
 
    protected:
-    /// Manages streans, cuBLAS handles and scratch memory for devices
+    /// Manages streams, cuBLAS handles and scratch memory for devices
     std::shared_ptr<GpuResources> resources_;
 
     /// Configuration options

faiss/gpu/GpuIndexBinaryFlat.h

@@ -86,7 +86,7 @@ class GpuIndexBinaryFlat : public IndexBinary {
             idx_t* outIndicesData) const;
 
    protected:
-    /// Manages streans, cuBLAS handles and scratch memory for devices
+    /// Manages streams, cuBLAS handles and scratch memory for devices
     std::shared_ptr<GpuResources> resources_;
 
     /// Configuration options

faiss/gpu/GpuIndexCagra.h

@@ -135,7 +135,7 @@ struct IVFPQSearchCagraConfig {
     ///
     /// The use of low-precision types reduces the amount of shared memory
     /// required at search time, so fast shared memory kernels can be used even
-    /// for datasets with large dimansionality. Note that the recall is slightly
+    /// for datasets with large dimensionality. Note that the recall is slightly
     /// degraded when low-precision type is selected.
 
     cudaDataType_t lut_dtype = CUDA_R_32F;

faiss/utils/distances.h

@@ -389,7 +389,7 @@ void knn_inner_products_by_idx(
  * @param x    query vectors, size nx * d
  * @param y    database vectors, size (max(ids) + 1) * d
  * @param subset subset of database vectors to consider, size (nx, nsubset)
- * @param res  rIDesult structure
+ * @param res  result structure
  * @param ld_subset stride for the subset array. -1: use nsubset, 0: all queries
  * process the same subset
  */