biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Add full API docs and improve the UX of navigating them (#17485)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-05-04 03:42:43 +01:00
committed by GitHub
parent 46fae69cf0
commit d6484ef3c3
101 changed files with 872 additions and 980 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/v1/worker/gpu_worker.py
View File

@@ -170,9 +170,10 @@ class Worker(WorkerBase):
Then, it calculate the free memory that can be used for KV cache in
bytes.
.. tip::
You may limit the usage of GPU memory
by adjusting the `gpu_memory_utilization` parameter.
:::{tip}
You may limit the usage of GPU memory
by adjusting the `gpu_memory_utilization` parameter.
:::
"""
torch.cuda.empty_cache()
torch.cuda.reset_peak_memory_stats()

6
vllm/v1/worker/utils.py
View File

@@ -10,7 +10,7 @@ def sanity_check_mm_encoder_outputs(
) -> None:
"""
Perform sanity checks for the result of
:meth:`vllm.model_executor.models.SupportsMultiModal.get_multimodal_embeddings`.
{meth}`vllm.model_executor.models.SupportsMultiModal.get_multimodal_embeddings`.
"""
assert isinstance(mm_embeddings, (list, tuple, torch.Tensor)), (
"Expected multimodal embeddings to be a list/tuple of 2D tensors, "
@@ -39,7 +39,7 @@ def scatter_mm_placeholders(
Scatter the multimodal embeddings into a contiguous tensor that represents
the placeholder tokens.
:class:`vllm.multimodal.processing.PromptUpdateDetails.is_embed`.
{class}`vllm.multimodal.processing.PromptUpdateDetails.is_embed`.
Args:
embeds: The multimodal embeddings.
@@ -66,7 +66,7 @@ def gather_mm_placeholders(
"""
Reconstructs the embeddings from the placeholder tokens.
This is the operation of :func:`scatter_mm_placeholders`.
This is the operation of {func}`scatter_mm_placeholders`.
"""
if is_embed is None:
return placeholders