Spaces:
Running
Running
File size: 2,270 Bytes
a0e37e2 9433533 a0e37e2 9433533 a0e37e2 c751e97 a0e37e2 9433533 a0e37e2 9433533 c751e97 9433533 a0e37e2 9433533 a0e37e2 9433533 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 |
from dataclasses import dataclass
from enum import Enum
import torch
from ask_candid.base.lambda_base import LambdaInvokeBase
@dataclass(slots=True)
class Encoding:
inputs: list[str]
vectors: torch.Tensor
@dataclass(slots=True)
class SummaryItem:
rank: int
score: float
text: str
@dataclass(slots=True)
class TextSummary:
snippets: list[SummaryItem]
@property
def summary(self) -> str:
return ' '.join([_.text for _ in self.snippets])
class CandidSLM(LambdaInvokeBase):
"""Wrapper around Candid's custom small language model.
For more details see https://dev.azure.com/guidestar/DataScience/_git/graph-ai?path=/releases/language.
This services includes:
* text encoding
* document summarization
* entity salience estimation
Parameters
----------
access_key : Optional[str], optional
AWS access key, by default None
secret_key : Optional[str], optional
AWS secret key, by default None
"""
class Tasks(Enum): # pylint: disable=missing-class-docstring
ENCODE = "/encode"
DOCUMENT_SUMMARIZE = "/document/summarize"
DOCUMENT_NER_SALIENCE = "/document/entitySalience"
def __init__(
self, access_key: str | None = None, secret_key: str | None = None
) -> None:
super().__init__(
function_name="small-lm",
access_key=access_key,
secret_key=secret_key
)
def encode(self, text: list[str]) -> Encoding:
response = self._submit_request({"text": text, "path": self.Tasks.ENCODE.value})
assert isinstance(response, dict)
return Encoding(
inputs=(response.get("inputs") or []),
vectors=torch.tensor((response.get("vectors") or []), dtype=torch.float32)
)
def summarize(self, text: list[str], top_k: int) -> TextSummary:
response = self._submit_request({"text": text, "path": self.Tasks.DOCUMENT_SUMMARIZE.value})
assert isinstance(response, dict)
return TextSummary(
snippets=[
SummaryItem(rank=item["rank"], score=item["score"], text=item["value"])
for item in (response.get("summary") or [])[:top_k]
]
)
|