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
@@ -179,6 +179,7 @@ class GenerateBaseServing(BaseServing, BeamSearchOnlineMixin):
|
||||
|
||||
Args:
|
||||
n: Number of sequences the request will occupy.
|
||||
|
||||
"""
|
||||
if self.engine_client.errored:
|
||||
raise self.engine_client.dead_error
|
||||
@@ -236,7 +237,7 @@ class GenerateBaseServing(BaseServing, BeamSearchOnlineMixin):
|
||||
|
||||
@staticmethod
|
||||
def _get_data_parallel_rank(raw_request: Request | None) -> int | None:
|
||||
"""Pulls the data parallel rank from a header, if provided"""
|
||||
"""Pulls the data parallel rank from a header, if provided."""
|
||||
if raw_request is None:
|
||||
return None
|
||||
|
||||
@@ -294,7 +295,8 @@ class GenerateBaseServing(BaseServing, BeamSearchOnlineMixin):
|
||||
) -> _T:
|
||||
"""Wrap a `create_*` coroutine so that, if it raises or returns an
|
||||
ErrorResponse (i.e. the request never reached the engine), the KV
|
||||
connector is notified to free any pinned remote-prefill blocks."""
|
||||
connector is notified to free any pinned remote-prefill blocks.
|
||||
"""
|
||||
kv_transfer_params = self.has_kv_connector and request.kv_transfer_params
|
||||
if not kv_transfer_params or not kv_transfer_params.get("do_remote_prefill"):
|
||||
return await awaitable
|
||||
|
||||
@@ -53,7 +53,7 @@ def _bitmask_to_token_ids(bitmask_row: torch.Tensor, vocab_size: int) -> list[in
|
||||
|
||||
|
||||
class BeamSearchOfflineMixin(OfflineInferenceMixin):
|
||||
"""Offline inference for beam search"""
|
||||
"""Offline inference for beam search."""
|
||||
|
||||
def beam_search(
|
||||
self,
|
||||
@@ -63,8 +63,7 @@ class BeamSearchOfflineMixin(OfflineInferenceMixin):
|
||||
use_tqdm: bool = False,
|
||||
concurrency_limit: int | None = None,
|
||||
) -> list[BeamSearchOutput]:
|
||||
"""
|
||||
Generate sequences using beam search.
|
||||
"""Generate sequences using beam search.
|
||||
|
||||
Args:
|
||||
prompts: A list of prompts. Each prompt can be a string or a list
|
||||
@@ -74,6 +73,7 @@ class BeamSearchOfflineMixin(OfflineInferenceMixin):
|
||||
use_tqdm: Whether to use tqdm to display the progress bar.
|
||||
concurrency_limit: The maximum number of concurrent requests.
|
||||
If None, the number of concurrent requests is unlimited.
|
||||
|
||||
"""
|
||||
# TODO: how does beam search work together with length penalty,
|
||||
# frequency, penalty, and stopping criteria, etc.?
|
||||
|
||||
@@ -20,7 +20,7 @@ from .utils import BeamSearchSequence, create_sort_beams_key_function
|
||||
|
||||
|
||||
class BeamSearchOnlineMixin(ABC):
|
||||
"""online serving for beam search"""
|
||||
"""online serving for beam search."""
|
||||
|
||||
renderer: BaseRenderer
|
||||
engine_client: EngineClient
|
||||
|
||||
@@ -60,6 +60,7 @@ class GenerativeScoringRequest(OpenAIBaseModel):
|
||||
the full vocab for those ids (False).
|
||||
item_first: If True, prepend items to query. Otherwise append items to query.
|
||||
add_special_tokens: Whether to add special tokens when tokenizing.
|
||||
|
||||
"""
|
||||
|
||||
model: str | None = None
|
||||
@@ -112,6 +113,7 @@ class GenerativeScoringItemResult(OpenAIBaseModel):
|
||||
index: The index of this item in the input items list.
|
||||
object: Type of object, always "score".
|
||||
score: The probability score for the first label token.
|
||||
|
||||
"""
|
||||
|
||||
index: int
|
||||
@@ -129,6 +131,7 @@ class GenerativeScoringResponse(OpenAIBaseModel):
|
||||
model: The model used for scoring.
|
||||
data: List of scoring results, one per input item.
|
||||
usage: Token usage information.
|
||||
|
||||
"""
|
||||
|
||||
id: str = Field(default="")
|
||||
@@ -188,6 +191,7 @@ class ServingGenerativeScoring(BaseServing):
|
||||
Returns:
|
||||
GenerativeScoringResponse with probabilities for each item, or
|
||||
ErrorResponse if an error occurred.
|
||||
|
||||
"""
|
||||
# Check model
|
||||
error_check_ret = await self._check_model(request) # type: ignore[arg-type]
|
||||
@@ -401,6 +405,7 @@ class ServingGenerativeScoring(BaseServing):
|
||||
|
||||
Returns:
|
||||
Tuple of (list of EngineInput, list of prompt token counts).
|
||||
|
||||
"""
|
||||
# Tokenize query if it's a string
|
||||
if isinstance(request.query, str):
|
||||
@@ -455,6 +460,7 @@ class ServingGenerativeScoring(BaseServing):
|
||||
|
||||
Returns:
|
||||
Dictionary mapping token_id to probability.
|
||||
|
||||
"""
|
||||
if apply_softmax:
|
||||
# Normalize over only the label tokens (subset softmax)
|
||||
|
||||
Reference in New Issue
Block a user