pymllm.engine.io_struct

Classes

Module Contents

class pymllm.engine.io_struct.BaseReq
rid: str | List[str] | None = None
regenerate_rid()
Return type:

Union[str, List[str]]

class pymllm.engine.io_struct.BaseBatchReq
rids: List[str]
regenerate_rids()
Return type:

List[str]

class pymllm.engine.io_struct.GenerateReqInput

Bases: BaseReq

text: List[str] | str | None = None
input_ids: List[List[int]] | List[int] | None = None
sampling_params: List[Dict[str, Any]] | Dict[str, Any] | None = None
return_logprob: List[bool] | bool | None = None
logprob_start_len: List[int] | int | None = None
top_logprobs_num: List[int] | int | None = None
stream: bool = False
image_data: Any | None = None
video_data: Any | None = None
audio_data: Any | None = None
lora_path: List[str | None] | str | None = None
session_params: List[Dict[str, Any]] | Dict[str, Any] | None = None
extra_options: Dict[str, Any]
is_single: bool = True
batch_size: int = 1
normalize_batch_and_arguments()
Return type:

None

__getitem__(i)
Parameters:

i (int)

Return type:

GenerateReqInput

to_request_dict()
Return type:

Dict[str, Any]

class pymllm.engine.io_struct.TokenizedGenerateReqInput

Bases: BaseReq

input_text: str = ''
input_ids: List[int] = []
mm_inputs: Dict[str, Any] | None = None
sampling_params: Dict[str, Any]
stream: bool = False
return_logprob: bool = False
logprob_start_len: int = -1
top_logprobs_num: int = 0
lora_path: str | None = None
session_params: Dict[str, Any] | None = None
class pymllm.engine.io_struct.BatchTokenizedGenerateReqInput

Bases: BaseBatchReq

reqs: List[TokenizedGenerateReqInput]
__len__()
Return type:

int

__getitem__(i)
Parameters:

i (int)

Return type:

TokenizedGenerateReqInput

__iter__()
Return type:

Iterator[TokenizedGenerateReqInput]

class pymllm.engine.io_struct.BatchTokenIDOutput

Bases: BaseBatchReq

finished_reasons: List[str | None]
decode_ids: List[int]
read_offsets: List[int]
output_ids: List[int] | None
skip_special_tokens: List[bool]
prompt_tokens: List[int]
completion_tokens: List[int]
input_token_logprobs_val: List[float] = []
input_token_logprobs_idx: List[int] = []
output_token_logprobs_val: List[float] = []
output_token_logprobs_idx: List[int] = []
input_top_logprobs_val: List[List[float]] = []
input_top_logprobs_idx: List[List[int]] = []
output_top_logprobs_val: List[List[float]] = []
output_top_logprobs_idx: List[List[int]] = []
class pymllm.engine.io_struct.BatchStrOutput

Bases: BaseBatchReq

finished_reasons: List[str | None]
output_strs: List[str]
output_ids: List[int] | None
prompt_tokens: List[int]
completion_tokens: List[int]
input_token_logprobs_val: List[float] = []
input_token_logprobs_idx: List[int] = []
output_token_logprobs_val: List[float] = []
output_token_logprobs_idx: List[int] = []
input_top_logprobs_val: List[List[float]] = []
input_top_logprobs_idx: List[List[int]] = []
output_top_logprobs_val: List[List[float]] = []
output_top_logprobs_idx: List[List[int]] = []