xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-21 23:15:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Docs] Fix warnings in mkdocs build (continued) (#25042)

Browse Source 
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wenlong Wang 2025-09-20 04:45:18 -07:00 committed by yewentao256
parent c2fdc71c91
commit dad5f4d16d
7 changed files with 24 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/multimodal/__init__.py
View File

@ -15,7 +15,7 @@ is used by model runners to dispatch data processing according to the target
model. model.
Info: Info:
[mm_processing](../../../design/mm_processing.html) [mm_processing](../../../design/mm_processing.md)
""" """
__all__ = [ __all__ = [

4
vllm/utils/__init__.py
View File

@ -3273,7 +3273,7 @@ def check_use_alibi(model_config: ModelConfig) -> bool:
and getattr(cfg.attn_config, "alibi", False))))) and getattr(cfg.attn_config, "alibi", False)))))
def sha256(input) -> bytes: def sha256(input: Any) -> bytes:
"""Hash any picklable Python object using SHA-256. """Hash any picklable Python object using SHA-256.
The input is serialized using pickle before hashing, which allows The input is serialized using pickle before hashing, which allows
@ -3290,7 +3290,7 @@ def sha256(input) -> bytes:
return hashlib.sha256(input_bytes).digest() return hashlib.sha256(input_bytes).digest()
def sha256_cbor(input) -> bytes: def sha256_cbor(input: Any) -> bytes:
""" """
Hash objects using CBOR serialization and SHA-256. Hash objects using CBOR serialization and SHA-256.

3
vllm/v1/worker/utils.py
View File

@ -205,7 +205,8 @@ def gather_mm_placeholders(
""" """
Reconstructs the embeddings from the placeholder tokens. Reconstructs the embeddings from the placeholder tokens.
This is the operation of [scatter_mm_placeholders][]. This is the operation of [`scatter_mm_placeholders`]
[vllm.v1.worker.utils.scatter_mm_placeholders].
""" """
if is_embed is None: if is_embed is None:
return placeholders return placeholders

12
vllm/worker/model_runner.py
View File

@ -1810,7 +1810,8 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
return [output] return [output]
def need_recv_kv(self, model_input, kv_caches) -> bool: def need_recv_kv(self, model_input: ModelInputForGPUWithSamplingMetadata,
kv_caches: List[torch.Tensor]) -> bool:
"""Check if we need to receive kv-cache from the other worker. """Check if we need to receive kv-cache from the other worker.
We need to receive KV when We need to receive KV when
1. current vLLM instance is KV cache consumer/decode vLLM instance 1. current vLLM instance is KV cache consumer/decode vLLM instance
@ -1825,6 +1826,9 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
if self.vllm_config.kv_transfer_config is None: if self.vllm_config.kv_transfer_config is None:
return False return False
if model_input.attn_metadata is None:
raise ValueError("model_input.attn_metadata cannot be None")
prefill_meta = model_input.attn_metadata.prefill_metadata prefill_meta = model_input.attn_metadata.prefill_metadata
# check if the current run is profiling # check if the current run is profiling
@ -1835,7 +1839,8 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
return self.vllm_config.kv_transfer_config.is_kv_consumer and ( return self.vllm_config.kv_transfer_config.is_kv_consumer and (
not is_profile_run) and is_prefill_run not is_profile_run) and is_prefill_run
def need_send_kv(self, model_input, kv_caches) -> bool: def need_send_kv(self, model_input: ModelInputForGPUWithSamplingMetadata,
kv_caches: List[torch.Tensor]) -> bool:
"""Check if we need to send kv-cache to the other worker. """Check if we need to send kv-cache to the other worker.
We need to send KV when We need to send KV when
1. current vLLM instance is KV cache producer/prefill vLLM instance 1. current vLLM instance is KV cache producer/prefill vLLM instance
@ -1850,6 +1855,9 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
if self.vllm_config.kv_transfer_config is None: if self.vllm_config.kv_transfer_config is None:
return False return False
if model_input.attn_metadata is None:
raise ValueError("model_input.attn_metadata cannot be None")
prefill_meta = model_input.attn_metadata.prefill_metadata prefill_meta = model_input.attn_metadata.prefill_metadata
# check if the current run is profiling # check if the current run is profiling