xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-09 03:15:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Doc]: fix typos in various files (#29230)

Browse Source 
Signed-off-by: Didier Durand <durand.didier@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Didier Durand 2025-11-24 12:10:48 +01:00 committed by GitHub
parent 8005e606bf
commit eca7a8fb59
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
benchmarks/kernels/deepgemm/README.md
View File

@ -2,7 +2,7 @@
This directory includes benchmarks between DeepSeek's DeepGEMM block fp8 kernels against vLLM's existing triton and CUTLASS-based kernels. This directory includes benchmarks between DeepSeek's DeepGEMM block fp8 kernels against vLLM's existing triton and CUTLASS-based kernels.
Currently this just includes dense GEMMs and only works on Hopper GPUs. Currently, this just includes dense GEMMs and only works on Hopper GPUs.
## Setup ## Setup

2
vllm/config/vllm.py
View File

@ -96,7 +96,7 @@ class VllmConfig:
"""`torch.compile` and cudagraph capture configuration for the model. """`torch.compile` and cudagraph capture configuration for the model.
As a shorthand, one can append compilation arguments via As a shorthand, one can append compilation arguments via
-0.parameter=arguement such as `-O.mode=3` (same as `-O='{"mode":3}'`). -0.parameter=argument such as `-O.mode=3` (same as `-O='{"mode":3}'`).
You can specify the full compilation config like so: You can specify the full compilation config like so:
`{"mode": 3, "cudagraph_capture_sizes": [1, 2, 4, 8]}` `{"mode": 3, "cudagraph_capture_sizes": [1, 2, 4, 8]}`

2
vllm/forward_context.py
View File

@ -153,7 +153,7 @@ class DPMetadata:
@contextmanager @contextmanager
def sp_local_sizes(self, sequence_parallel_size: int): def sp_local_sizes(self, sequence_parallel_size: int):
""" """
Context mamager for setting self.local_sizes. Same as self.chunked_sizes Context manager for setting self.local_sizes. Same as self.chunked_sizes
but without any chunking. but without any chunking.
""" """
self.local_sizes = _compute_sp_num_tokens( self.local_sizes = _compute_sp_num_tokens(

2
vllm/v1/worker/gpu_input_batch.py
View File

@ -525,7 +525,7 @@ class InputBatch:
# NOTE: the following is unsafe # NOTE: the following is unsafe
# self.token_ids_cpu[i1, ...], self.token_ids_cpu[i2, ...], =\ # self.token_ids_cpu[i1, ...], self.token_ids_cpu[i2, ...], =\
# self.token_ids_cpu[i2, ...], self.token_ids_cpu[i1, ...] # self.token_ids_cpu[i2, ...], self.token_ids_cpu[i1, ...]
# instead, we need to temporiarily copy the data for one of the indices # instead, we need to temporarily copy the data for one of the indices
# TODO(lucas): optimize this by only copying valid indices # TODO(lucas): optimize this by only copying valid indices
tmp = self.token_ids_cpu[i1, ...].copy() tmp = self.token_ids_cpu[i1, ...].copy()
self.token_ids_cpu[i1, ...] = self.token_ids_cpu[i2, ...] self.token_ids_cpu[i1, ...] = self.token_ids_cpu[i2, ...]