Fixed typo: begining -> beginning

codegeex/megatron/data/helpers.cpp

@@ -158,7 +158,7 @@ py::array build_sample_idx(const py::array_t<int32_t>& sizes_,
 	  doc_offset += (remaining_seq_length + doc_length - 1);
 	  remaining_seq_length = 0;
 	} else {
-	  // Otherwise, start from the begining of the next document.
+	  // Otherwise, start from the beginning of the next document.
 	  ++doc_idx_index;
 	  doc_offset = 0;
 	}
@@ -296,7 +296,7 @@ py::array build_mapping_impl(const py::array_t<int64_t>& docs_,
                 const auto sent_index_first = docs[doc];
                 const auto sent_index_last = docs[doc + 1];
 
-                // At the begining of the document previous index is the
+                // At the beginning of the document previous index is the
 		// start index.
                 auto prev_start_index = sent_index_first;
 
@@ -556,7 +556,7 @@ py::array build_blocks_mapping_impl(const py::array_t<int64_t>& docs_,
                 const auto sent_index_last = docs[doc + 1];
                 const auto target_seq_len = max_seq_length - titles_sizes[doc];
 
-                // At the begining of the document previous index is the
+                // At the beginning of the document previous index is the
                 // start index.
                 auto prev_start_index = sent_index_first;
 

codegeex/megatron/model/distributed.py

@@ -174,7 +174,7 @@ class DistributedDataParallel(DistributedDataParallelBase):
 
     def zero_grad_buffer(self):
         """Set the grad buffer data to zero. Needs to be called at the
-        begining of each iteration."""
+        beginning of each iteration."""
         assert self._grad_buffers is not None, "buffers are not initialized."
         for _, buffer_ in self._grad_buffers.items():
             buffer_.zero()

codegeex/megatron/model/transformer.py

@@ -779,7 +779,7 @@ class ParallelTopQueryLayer(MegatronModule):
         # hidden_states: [b, s, h]
         assert query_hidden_state != None
 
-        # Layer norm at the begining of the transformer layer.
+        # Layer norm at the beginning of the transformer layer.
         if self.ln_fp16:
             layernorm_output = self.input_layernorm(hidden_states)
         else: