Add pydocstyle to the ruff rules (#52136)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Signed-off-by: Harry Mellor <hej.mellor@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
804e5377aa
commit
8c1557a79c
@@ -1,7 +1,6 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
"""
|
||||
Triton kernels for DeepseekV4 paged K-cache management and sparse-attention index
|
||||
"""Triton kernels for DeepseekV4 paged K-cache management and sparse-attention index
|
||||
preparation.
|
||||
|
||||
- quantize_and_insert_k_cache: quantize bf16 K to UE8M0 FP8 and insert into
|
||||
@@ -61,8 +60,7 @@ def quantize_and_insert_k_kernel(
|
||||
n_quant_blocks: tl.constexpr, # 8 (7 real + 1 padding)
|
||||
use_fnuz: tl.constexpr = False,
|
||||
):
|
||||
"""
|
||||
Quantize K tensor and insert into paged K cache.
|
||||
"""Quantize K tensor and insert into paged K cache.
|
||||
|
||||
K Cache block layout (block_size=64 tokens):
|
||||
- [0, 64*576): Token data, each token has 448 fp8 + 128 bf16
|
||||
@@ -174,8 +172,7 @@ def quantize_and_insert_k_cache(
|
||||
is_ue8m0: bool = True,
|
||||
use_fnuz: bool = False,
|
||||
):
|
||||
"""
|
||||
Quantize K tensor and insert into paged K cache.
|
||||
"""Quantize K tensor and insert into paged K cache.
|
||||
|
||||
K Cache block layout (block_size=64 tokens):
|
||||
- First 64 * 576 = 36864 bytes: Token data
|
||||
|
||||
Reference in New Issue
Block a user