[VLM][Bugfix] Make sure that multi_modal_kwargs is broadcasted properly (#5880)

Signed-off-by: Xiaowei Jiang <[email protected]>

Author: xwjiang2010

.buildkite/test-pipeline.yaml

@@ -27,7 +27,9 @@ steps:
 
 - label: Core Test
   mirror_hardwares: [amd]
-  command: pytest -v -s core
+  commands: 
+  - pytest -v -s core
+  - pytest -v -s distributed/test_parallel_state.py
 
 - label: Distributed Comm Ops Test
   #mirror_hardwares: [amd]

tests/distributed/test_parallel_state.py

@@ -0,0 +1,49 @@
+from typing import Any, Dict
+
+import torch
+
+from vllm.distributed.parallel_state import (_split_tensor_dict,
+                                             _update_nested_dict)
+
+
+def test_split_tensor_dict():
+    test_dict = {
+        "key_a": "a",
+        "key_b": torch.arange(8, dtype=torch.float32),
+        "key_c": {
+            "key_1": torch.arange(5, dtype=torch.float32),
+            "key_2": torch.tensor([], dtype=torch.float32),
+            "key_3": 123,
+        },
+        "key_d": {},
+    }
+    metadata_list, tensor_list = _split_tensor_dict(test_dict)
+    assert len(metadata_list) == 6
+    assert torch.allclose(tensor_list[0], test_dict["key_b"])
+    assert torch.allclose(tensor_list[1], test_dict["key_c"]["key_1"])
+    assert torch.allclose(tensor_list[2], test_dict["key_c"]["key_2"])
+
+
+def test_update_nested_dict():
+    flattened_keys_values = [("key1%key2%key3", "value1"),
+                             ("key1%key2%key4", "value2"),
+                             ("key1%key5", "value3"), ("key6%key7", "value4"),
+                             ("key8", "value5")]
+    res: Dict[str, Any] = {}
+
+    # Update the nested dictionary with each flattened key-value pair
+    for flat_key, value in flattened_keys_values:
+        _update_nested_dict(res, flat_key, value)
+    assert res == {
+        "key1": {
+            "key2": {
+                "key3": "value1",
+                "key4": "value2"
+            },
+            "key5": "value3"
+        },
+        "key6": {
+            "key7": "value4"
+        },
+        "key8": "value5"
+    }

vllm/distributed/parallel_state.py

@@ -45,14 +45,17 @@ class GraphCaptureContext:
 
 
 def _split_tensor_dict(
-    tensor_dict: Dict[Any, Union[torch.Tensor, Any]]
-) -> Tuple[List[Tuple[str, Any]], List[torch.Tensor]]:
+        tensor_dict: Dict[Any, Union[torch.Tensor, Any]],
+        prefix: str = "") -> Tuple[List[Tuple[str, Any]], List[torch.Tensor]]:
     """Split the tensor dictionary into two parts:
     1. A list of (key, value) pairs. If the value is a tensor, it is replaced
          by its metadata.
     2. A list of tensors.
+
+    If the Tensor is nested under `tensor_dict["key1"]["key2"]`, the key of its
+    metadata will be "key1%key2".
     """
-    metadata_list = []
+    metadata_list: List[Tuple[str, Any]] = []
     tensor_list = []
     for key, value in tensor_dict.items():
         if isinstance(value, torch.Tensor):
@@ -62,13 +65,31 @@ def _split_tensor_dict(
             # receiving side will set the device index.
             device = value.device.type
             metadata_list.append(
-                (key, TensorMetadata(device, value.dtype, value.size())))
+                (prefix + key, TensorMetadata(device, value.dtype,
+                                              value.size())))
             tensor_list.append(value)
+        elif isinstance(value, dict):
+            if len(value) == 0:
+                metadata_list.append((prefix + key, value))
+            inner_metadata_list, inner_tensor_list = _split_tensor_dict(
+                value, prefix + key + "%")
+            metadata_list.extend(inner_metadata_list)
+            tensor_list.extend(inner_tensor_list)
         else:
-            metadata_list.append((key, value))
+            metadata_list.append((prefix + key, value))
     return metadata_list, tensor_list
 
 
+def _update_nested_dict(nested_dict, flattened_key, value):
+    key_splits = flattened_key.split("%")
+    cur_dict = nested_dict
+    for k in key_splits[:-1]:
+        if k not in cur_dict:
+            cur_dict[k] = {}
+        cur_dict = cur_dict[k]
+    cur_dict[key_splits[-1]] = value
+
+
 class GroupCoordinator:
     """
     PyTorch ProcessGroup wrapper for a group of processes.
@@ -512,7 +533,7 @@ def broadcast_tensor_dict(
                                          device=value.device)
                     if tensor.numel() == 0:
                         # Skip broadcasting empty tensors.
-                        tensor_dict[key] = tensor
+                        _update_nested_dict(tensor_dict, key, tensor)
                         continue
                     if tensor.is_cpu:
                         # use metadata_group for CPU tensors
@@ -528,9 +549,9 @@ def broadcast_tensor_dict(
                                                              group=group,
                                                              async_op=True)
                     async_handles.append(handle)
-                    tensor_dict[key] = tensor
+                    _update_nested_dict(tensor_dict, key, tensor)
                 else:
-                    tensor_dict[key] = value
+                    _update_nested_dict(tensor_dict, key, value)
             for async_handle in async_handles:
                 async_handle.wait()
         return tensor_dict