[Refactor] Add comments for Metadata classes in attention module (#5789)

### What this PR does / why we need it?

Add docstrings for Metadata and MetadataBuilder classes in the attention
module to improve code readability.

Related to #5463 (Item 11: Add some comments for CommonMetadata and
others)

**Modified files:**
- `vllm_ascend/attention/context_parallel/common_cp.py`: Added comments
for `AscendPCPMetadata`, `CPChunkedContextMetadata`,
`AscendMetadataForPrefill`, `AscendMetadataForDecode`
- `vllm_ascend/attention/utils.py`: Added comments for
`AscendPrefillContextParallelMetadata`
- `vllm_ascend/attention/mla_v1.py`: Added comments for
`ChunkedContextMetadata`, `AscendMLADecodeMetadata`
- `vllm_ascend/attention/attention_v1.py`: Added comments for
`AscendMetadata`, `AscendAttentionMetadataBuilder`
- `vllm_ascend/attention/context_parallel/attention_cp.py`: Added
comments for `AscendAttentionCPMetadataBuilder`

### Does this PR introduce _any_ user-facing change?

No.

### How was this patch tested?

Documentation only, no functional changes.

Signed-off-by: lico67373 <918688502@qq.com>
This commit is contained in:
LICO67373
2026-01-13 08:46:50 +08:00
committed by GitHub
parent dde547e900
commit c8a324ab73
5 changed files with 58 additions and 11 deletions

View File

@@ -84,8 +84,11 @@ class AscendMLABackend(AttentionBackend):
@dataclass
class ChunkedContextMetadata:
# New for MLA (compared to FlashAttention)
# For handling chunked prefill
"""
Metadata for chunked context handling in MLA attention.
Manages sequence boundaries and workspace for chunked prefill processing.
"""
cu_seq_lens: torch.Tensor
starts: torch.Tensor
seq_tot: list[int]
@@ -116,7 +119,8 @@ class AscendMLAPrefillMetadata:
@dataclass
class AscendMLADecodeMetadata:
# Input positions for rotrary embeddings since for MLA the rotary
""" Decode-specific metadata for Ascend MLA attention."""
# Input positions for rotary embeddings since for MLA the rotary
# position embeddings are applied inside the attention backend
input_positions: torch.Tensor
block_table: torch.Tensor