def setUp(self):
# setup logger
server.logger = utils_logging.getLogger(logging.DEBUG, "logs")
# random port
socket = server.sockets.PySocket(port=0)
# random datebase path name without dashes
database_path = "./{}.db".format(str(uuid.uuid4().hex))
self.s = server.AEPsychServer(socket=socket, database_path=database_path)
def setUp(self):
# setup logger
server.logger = utils_logging.getLogger(logging.DEBUG, "logs")
socket = server.sockets.ThriftSocketWrapper(msg_queue=Queue())
# random datebase path name without dashes
database_path = "./{}.db".format(str(uuid.uuid4().hex))
self.s = server.AEPsychServer(socket=socket,
database_path=database_path,
thrift=True)
from aepsych.utils import promote_0d, _dim_grid, get_jnd_multid
from botorch.acquisition import (
qNoisyExpectedImprovement,
qUpperConfidenceBound,
NoisyExpectedImprovement,
)
from scipy.stats import norm
# this is pretty aggressive jitter setting but should protect us from
# crashes which are a bigger concern in data collection.
# we refit with stan afterwards anyway.
gpytorch.settings.cholesky_jitter._global_float_value = 1e-3
gpytorch.settings.cholesky_jitter._global_double_value = 1e-3
gpytorch.settings.tridiagonal_jitter._global_value = 1e-3
logger = utils_logging.getLogger(logging.DEBUG)
def _prune_extra_acqf_args(acqf, extra_acqf_args):
# prune extra args needed, ignore the rest
# (this helps with API consistency)
acqf_args_expected = signature(acqf).parameters.keys()
return {k: v for k, v in extra_acqf_args.items() if k in acqf_args_expected}
def _prune_extra_acqf_args(acqf, extra_acqf_args):
# prune extra args needed, ignore the rest
# (this helps with API consistency)
acqf_args_expected = signature(acqf).parameters.keys()
return {k: v for k, v in extra_acqf_args.items() if k in acqf_args_expected}
# Copyright (c) Facebook, Inc. and its affiliates.
# All rights reserved.
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import json
import logging
import socket
import sys
import aepsych.utils_logging as utils_logging
import numpy as np
import zmq
logger = utils_logging.getLogger(logging.INFO)
def SimplifyArrays(message):
return {
k: v.tolist() if type(v) == np.ndarray else
SimplifyArrays(v) if type(v) is dict else v
for k, v in message.items()
}
def createSocket(socket_type="pysocket", port=5555, msg_queue=None):
logger.info(f"socket_type = {socket_type} port = {port}")
if socket_type == "pysocket":
sock = PySocket(port=port)
import torch
from aepsych.config import Config
from aepsych.factory.factory import default_mean_covar_factory
from aepsych.models.base import AEPsychMixin
from aepsych.utils import _process_bounds, promote_0d
from aepsych.utils_logging import getLogger
from botorch.fit import fit_gpytorch_model
from botorch.models.gpytorch import GPyTorchModel
from gpytorch.likelihoods import BernoulliLikelihood, Likelihood
from gpytorch.models import ApproximateGP
from gpytorch.variational import CholeskyVariationalDistribution, VariationalStrategy
from scipy.special import owens_t
from scipy.stats import norm
from torch.distributions import Normal
logger = getLogger()
class GPClassificationModel(AEPsychMixin, ApproximateGP, GPyTorchModel):
"""Probit-GP model with variational inference.
From a conventional ML perspective this is a GP Classification model,
though in the psychophysics context it can also be thought of as a
nonlinear generalization of the standard linear model for 1AFC or
yes/no trials.
For more on variational inference, see e.g.
https://docs.gpytorch.ai/en/v1.1.1/examples/04_Variational_and_Approximate_GPs/
"""
_num_outputs = 1