[Doc]: fixing doc typos (vllm-project#24635)

Signed-off-by: Didier Durand <durand.didier@gmail.com>

Author: didier-durand

vllm/config/__init__.py

@@ -3881,7 +3881,7 @@ def contains_object_print(text):
     Check if the text looks like a printed Python object, e.g.
     contains any substring matching the pattern: "at 0xFFFFFFF>"
     We match against 0x followed by 2-16 hex chars (there's
-    a max of 16 on a 64 bit system).
+    a max of 16 on a 64-bit system).
 
     Args:
         text (str): The text to check

vllm/entrypoints/openai/tool_parsers/internlm2_tool_parser.py

@@ -60,7 +60,7 @@ def extract_tool_calls_streaming(
         if '<|action_start|>' not in current_text:
             self.position = len(current_text)
             return DeltaMessage(content=delta_text)
-        # if the tool call is sended, return an empty delta message
+        # if the tool call is sent, return an empty delta message
         # to make sure the finish_reason will be sent correctly.
         if self.current_tool_id > 0:
             return DeltaMessage(content='')

vllm/model_executor/layers/mamba/ops/ssd_chunk_state.py

@@ -502,7 +502,7 @@ def _chunk_state_varlen_kernel(
         dA_cumsum_ptrs += BLOCK_SIZE_K * stride_dA_cs_csize
 
     # If the sequence starts after the last chunk idx, we don't need to add the contribution from the last chunk
-    # If HAS_INITSTATES==True need to consider two possiblties
+    # If HAS_INITSTATES==True need to consider two possibilities
     # - if start_idx < pid_c * chunk_size, then we need to take the past_states_ptrs
     # - if state_idx >= pid * chunk_size, then we need to insert initstates
     if ((start_idx < pid_c * chunk_size)  # first chunk

vllm/model_executor/models/arcee.py

@@ -342,7 +342,7 @@ def load_weights(self, weights: Iterable[tuple[str,
 class ArceeForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
     """Arcee Model for causal language modeling, integrated with vLLM
     runtime."""
-    # Map fused module names to their sub-module components
+    # Map fused module names to their submodule components
     # (for quantization and LoRA)
     packed_modules_mapping = {
         "qkv_proj": ["q_proj", "k_proj", "v_proj"],

vllm/model_executor/models/llava_onevision.py

@@ -835,7 +835,7 @@ def get_multimodal_embeddings(self,
             return None
 
         # The result multimodal_embeddings is tuple of tensors, with each
-        # tensor correspoending to a multimodal data item (image or video).
+        # tensor corresponding to a multimodal data item (image or video).
         multimodal_embeddings: tuple[torch.Tensor, ...] = ()
 
         # NOTE: It is important to iterate over the keys in this dictionary

vllm/model_executor/models/phi4_multimodal.py

@@ -1350,7 +1350,7 @@ def get_multimodal_embeddings(
             return None
 
         # The result multimodal_embeddings is tuple of tensors, with each
-        # tensor correspoending to a multimodal data item (image or video).
+        # tensor corresponding to a multimodal data item (image or video).
         multimodal_embeddings: tuple[torch.Tensor, ...] = ()
 
         # NOTE: It is important to iterate over the keys in this dictionary

vllm/model_executor/models/phi4mm_audio.py

@@ -100,7 +100,7 @@ class ConformerEncoderLayer(nn.Module):
             activation function for glu used in the multihead attention,
              default "swish".
         activation_checkpointing: str, optional
-            a dictionarry of {"module","interval","offload"}, where
+            a dictionary of {"module","interval","offload"}, where
                 "module": str
                     accept ["transformer", "attention"] to select
                     which module should do activation checkpointing.

vllm/model_executor/models/qwen2_5_omni_thinker.py

@@ -846,7 +846,7 @@ def get_multimodal_embeddings(self,
             return []
 
         # The result multimodal_embeddings is tuple of tensors, with each
-        # tensor correspoending to a multimodal data item (image or video).
+        # tensor corresponding to a multimodal data item (image or video).
         multimodal_embeddings: tuple[torch.Tensor, ...] = ()
 
         # NOTE: It is important to iterate over the keys in this dictionary
@@ -873,7 +873,7 @@ def get_input_embeddings(
         if multimodal_embeddings is not None \
             and len(multimodal_embeddings) != 0:
 
-            # TODO (ywang96): support overlapping modalitiy embeddings so that
+            # TODO (ywang96): support overlapping modality embeddings so that
             # `use_audio_in_video` will work on V1.
             inputs_embeds = merge_multimodal_embeddings(
                 input_ids, inputs_embeds, multimodal_embeddings, [

vllm/v1/attention/backends/mla/common.py

@@ -463,7 +463,7 @@ def __init__(self,
             self.dcp_world_size = 1
             self.dcp_rank = 0
 
-        # Dont try to access the runner on AMD
+        # Don't try to access the runner on AMD
         if self.aot_schedule:
             self.page_size = self.kv_cache_spec.block_size