easydel.inference.esurge.scheduler.output

class easydel.inference.esurge.scheduler.output.CachedRequestData(req_ids: 'list[str]', resumed_from_preemption: 'list[bool]', new_token_ids: 'list[list[int]]', new_page_ids: 'list[tuple[list[int], ...]]', num_computed_tokens: 'list[int]')

Bases: object

classmethod make_empty() → CachedRequestData

new_page_ids: list[tuple[list[int], ...]]

new_token_ids: list[list[int]]

num_computed_tokens: list[int]

property num_reqs: int

req_ids: list[str]

resumed_from_preemption: list[bool]

class easydel.inference.esurge.scheduler.output.NewRequestData(req_id: 'str', prompt_token_ids: 'list[int]', sampling_params: 'SamplingParams | None', page_ids: 'tuple[list[int], ...]', num_computed_tokens: 'int')

Bases: object

anon_repr()

classmethod from_request(request: EngineRequest, page_ids: tuple[list[int], ...]) → NewRequestData

num_computed_tokens: int

page_ids: tuple[list[int], ...]

prompt_token_ids: list[int]

req_id: str

sampling_params: SamplingParams | None

class easydel.inference.esurge.scheduler.output.SchedulerOutput(scheduled_new_reqs: 'list[NewRequestData]', scheduled_cached_reqs: 'CachedRequestData', num_scheduled_tokens: 'dict[str, int]', total_num_scheduled_tokens: 'int', scheduled_spec_decode_tokens: 'dict[str, list[int]]', num_common_prefix_pages: 'list[int]', finished_req_ids: 'set[str]', suggested_bucket: 'int | None' = None, async_scheduling: 'bool' = False)

Bases: object

async_scheduling: bool = False

finished_req_ids: set[str]

num_common_prefix_pages: list[int]

num_scheduled_tokens: dict[str, int]

scheduled_cached_reqs: CachedRequestData

scheduled_new_reqs: list[easydel.inference.esurge.scheduler.output.NewRequestData]

scheduled_spec_decode_tokens: dict[str, list[int]]

suggested_bucket: int | None = None

total_num_scheduled_tokens: int