[Doc]: fix typos in Python comments (#24115)

Signed-off-by: Didier Durand <durand.didier@gmail.com>
This commit is contained in:
Didier Durand
2025-09-02 21:14:07 -07:00
committed by GitHub
parent d7e1e59972
commit 02d411fdb2
11 changed files with 11 additions and 11 deletions
+1 -1
View File
@@ -513,7 +513,7 @@ if flashinfer_comm is not None:
torch.ops._C.static_scaled_fp8_quant(
quant_out, norm_out, scale_factor)
if scale_factor is None or norm_out is not None:
# we need to return allreduce outpput
# we need to return allreduce output
# in cases of non quant fused AR + RMS norm
# and fused AR + RMS norm + quant without fused add
allreduce_in.copy_(allreduce_out)
+1 -1
View File
@@ -49,7 +49,7 @@ class MQLLMEngine:
This class is used to wrap the
[`LLMEngine`][vllm.engine.llm_engine.LLMEngine] class to enable use
in concurrnet manner. It runs a background loop and uses zeromq to
in concurrent manner. It runs a background loop and uses zeromq to
receive new requests and stream outputs incrementally via ipc.
The [`LLMEngine`][vllm.engine.llm_engine.LLMEngine] generate or encode
@@ -23,7 +23,7 @@ TORCH_DEVICE_IDENTITY = None
# The condition to determine if it is on a platform that supports
# torch._scaled_mm rowwise feature.
# The condition is determined once as the operations
# are time consuming.
# are time-consuming.
USE_ROWWISE_TORCH_SCALED_MM = (current_platform.is_rocm() and version.parse(
torch.__version__) >= version.parse("2.7")
and current_platform.has_device_capability(94))
@@ -211,7 +211,7 @@ class DefaultModelLoader(BaseModelLoader):
if not USE_TPU_COMMONS:
# In PyTorch XLA, we should call `xm.mark_step`
# requently so that not too many ops are accumulated
# frequently so that not too many ops are accumulated
# in the XLA program. import torch_xla.core.xla_model
# as xm
import torch_xla.core.xla_model as xm
+1 -1
View File
@@ -84,7 +84,7 @@ class XPUWorker(Worker):
"""Profiles the peak memory usage of the model to determine how many
KV blocks may be allocated without OOMs.
The engine will first conduct a profiling of the existing memory usage.
Then, it calculate the maximum possible number of GPU and CPU blocks
Then, it calculates the maximum possible number of GPU and CPU blocks
that can be allocated with the remaining free memory.
.. tip::
You may limit the usage of GPU memory
+1 -1
View File
@@ -234,7 +234,7 @@ class Worker(LocalOrDistributedWorkerBase):
KV blocks may be allocated without OOMs.
The engine will first conduct a profiling of the existing memory usage.
Then, it calculate the maximum possible number of GPU and CPU blocks
Then, it calculates the maximum possible number of GPU and CPU blocks
that can be allocated with the remaining free memory.
Tip: