63 lines
3.2 KiB
Python
63 lines
3.2 KiB
Python
from typing import Tuple
|
|
|
|
from flask import jsonify
|
|
from vllm import SamplingParams
|
|
|
|
from llm_server.database.log_to_db import log_to_db
|
|
from llm_server.llm.llm_backend import LLMBackend
|
|
|
|
|
|
class VLLMBackend(LLMBackend):
|
|
_default_params = vars(SamplingParams())
|
|
|
|
def handle_response(self, success, request, response_json_body, response_status_code, client_ip, token, prompt: str, elapsed_time, parameters, headers):
|
|
if len(response_json_body.get('text', [])):
|
|
# Does vllm return the prompt and the response together???
|
|
backend_response = response_json_body['text'][0].split(prompt)[1].strip(' ').strip('\n')
|
|
else:
|
|
# Failsafe
|
|
backend_response = ''
|
|
|
|
log_to_db(ip=client_ip, token=token, prompt=prompt, response=backend_response, gen_time=elapsed_time, parameters=parameters, headers=headers, backend_response_code=response_status_code, request_url=request.url,
|
|
response_tokens=response_json_body.get('details', {}).get('generated_tokens'), backend_url=self.backend_url)
|
|
|
|
return jsonify({'results': [{'text': backend_response}]}), 200
|
|
|
|
def get_parameters(self, parameters) -> Tuple[dict | None, str | None]:
|
|
"""
|
|
Convert the Oobabooga parameters to VLLM and validate them.
|
|
:param parameters:
|
|
:return:
|
|
"""
|
|
try:
|
|
# top_k == -1 means disabled
|
|
top_k = parameters.get('top_k', self._default_params['top_k'])
|
|
if top_k <= 0:
|
|
top_k = -1
|
|
|
|
# We call the internal VLLM `SamplingParams` class to validate the input parameters.
|
|
# Parameters from Oobabooga don't line up here exactly, so we have to shuffle some things around.
|
|
# TODO: support more params
|
|
sampling_params = SamplingParams(
|
|
temperature=parameters.get('temperature', self._default_params['temperature']),
|
|
top_p=parameters.get('top_p', self._default_params['top_p']),
|
|
top_k=top_k,
|
|
use_beam_search=True if parameters.get('num_beams', 0) > 1 else False,
|
|
stop=list(set(parameters.get('stopping_strings') or parameters.get('stop', self._default_params['stop']))),
|
|
ignore_eos=parameters.get('ban_eos_token', False),
|
|
max_tokens=parameters.get('max_new_tokens') or parameters.get('max_tokens', self._default_params['max_tokens']),
|
|
presence_penalty=parameters.get('presence_penalty', self._default_params['presence_penalty']),
|
|
frequency_penalty=parameters.get('frequency_penalty', self._default_params['frequency_penalty']),
|
|
length_penalty=parameters.get('length_penalty', self._default_params['length_penalty']),
|
|
early_stopping=parameters.get('early_stopping', self._default_params['early_stopping'])
|
|
)
|
|
except ValueError as e:
|
|
# `SamplingParams` will return a pretty error message. Send that back to the caller.
|
|
return None, str(e).strip('.')
|
|
|
|
# We use max_new_tokens throughout this program, so rename the variable.
|
|
result = vars(sampling_params)
|
|
result['max_new_tokens'] = result.pop('max_tokens')
|
|
|
|
return result, None
|