""" Client gRPC services for running on FPGA reservoir computer EmuCore developed by QCi. """ import logging import json import time from typing import List , TypedDict import os import grpc from grpc._channel import _InactiveRpcError import numpy as np from . import emucore_pb2, emucore_pb2_grpc from .utils import set_filter_coefficients, \ bytes_to_array, \ prep_input_data, \ message_to_dict grpc_service_config = json.dumps( { "methodConfig" : [ { "name" : [{ "service" : "EmuCore.EmuCoreService" }], "retryPolicy" : { "maxAttempts" : 5 , "initialBackoff" : "0.2s" , "maxBackoff" : "10s" , "backoffMultiplier" : 2.5 , "retryableStatusCodes" : [ "UNAVAILABLE" ], }, } ] } ) MAX_INPUT_SIZE = 20 * 1024 * 1024 class StatusMessage ( TypedDict ): """ Structure of responses for configuration requests to EmuCore device. :param status: the status of the request :param message: a description for the recieved status """ status: int message: str class InactiveRpcError ( Exception ): """Custom exception wrapper around grpc._channel._InactiveRpcError.""" class EmuCoreClient : """ Provides services for accessing EmuCore server :param ip_addr: the IP address of the gRPC server :param port: The port that the RPC server is running on :param max_data_size: int The max send and recieve message length for RPC server .. note:: :code:`lock_id` is used by a variety of class functions. It is set to an empty string by default since default for device server :code:`lock_id` is also an empty string. This allows for single user processing without having to acquire a device lock. .. All GRPC calls follow a specific pattern: .. 1. Fill in data to be sent in message stub .. 2. Send data using stub service method .. 3. Parse response """ def __init__ ( self, ip_addr: str = os.getenv( "DEVICE_IP_ADDRESS" , "localhost" ), port: str = os.getenv( "DEVICE_PORT" , "50051" ), max_data_size: int = 512 * 1024 * 1024 , ): self ._ip_add_port = ip_addr + ":" + port self ._channel_opt = [ ( "grpc.max_send_message_length" , max_data_size), ( "grpc.max_receive_message_length" , max_data_size), ( "grpc.service_config" , grpc_service_config), ] self .channel = grpc.insecure_channel( self ._ip_add_port, options= self ._channel_opt ) self .stub = emucore_pb2_grpc.EmuCoreServiceStub( self .channel) def check_lock ( self, lock_id: str = "" ) -> dict : """ Checks if submitted :code:`lock_id` has execution lock on the device :param lock_id: a UUID which will be checked to determine if has exclusive device execution lock :return: a member of :class:`eqc_direct.utils.LockCheckStatus` as a dict: - **status_code**: `int`- status code for lock check - **status_desc**: `str`- a description for the associated status code """ check_input = emucore_pb2.lock_message(lock_id=lock_id) check_output = self .eqc_stub.check_lock(check_input) return message_to_dict(check_output) def reservoir_reset ( self, lock_id ) -> StatusMessage: """ Resets a reservoir instance by clearing RAM on the server :param lock_id: a lock_id which has an active reserve on the device :return: dictionary with with values from members of :class:`emucore_direct.types.StatusResponses` """ reset_message = emucore_pb2.lock_message(lock_id=lock_id) return message_to_dict( self .stub.reservoir_reset(reset_message)) def system_info ( self ) -> dict : """ Provides system info on call :return: dict with following keys: - **system_name**: `str`- product name - **system_version**: `str`- server version """ sys_info_resp = self .stub.system_info(emucore_pb2.empty_message()) return message_to_dict(sys_info_resp) def acquire_lock ( self ) -> dict : """ Acquires exclusive lock for running health check or submitting job :return: a member of :class:`emucore_direct.types.LockManageStatus` as a dict along with an additional key :code:`lock_id`: - **lock_id**: `str`- if acquired the current device `lock_id` else empty string - **status**: `int`- status code for lock id acquisition - **message**: `str`- a description for the associated status code """ try : acquire_lock_resp = self .stub.acquire_lock(emucore_pb2.empty_message()) except _InactiveRpcError as exc: raise InactiveRpcError( "acquire_lock failed due to grpc._channel._InactiveRpcError." ) from exc return message_to_dict(acquire_lock_resp) def release_lock ( self, lock_id ): """ Releases exclusive lock for running health check or submitting job :param lock_id: a UUID with currently acquired exclusive device lock :return: a dict with the following keys: - **lock_released**: `bool`- if released is True else False - **message**: `str`- a description of release operation result """ release_input = emucore_pb2.lock_message(lock_id = lock_id) try : release_lock_resp = self .stub.release_lock(release_input) except _InactiveRpcError as exc: raise InactiveRpcError( "release_lock failed due to grpc._channel._InactiveRpcError." ) from exc return message_to_dict(release_lock_resp) def rc_config ( self, lock_id: str , vbias: float , gain: float , num_nodes: int , num_taps: int ) -> StatusMessage: """ :param lock_id: a lock_id which has an active reserve on the device :param vbias: bias to apply to each node in reservoir range for parameter [0,1] :param gain: memory setting for system how long should inputs effect reservoir similar to beta in adaptive gradient descent range for parameter [0,1] :param num_nodes: the total number of hidden nodes to instantiate within the reservoir, a single hidden layer :param num_taps: number of connections in reservoir, generally should be set to less than the number of nodes in reservoir. Defines interconnection between nodes. :return: dictionary with with values from one of the members of :class:`emucore_direct.types.StatusResponses` """ filter_coefs = set_filter_coefficients(num_taps=num_taps) config_message = emucore_pb2.rc_config_message(lock_id=lock_id, vbias = vbias, gain = gain, num_nodes = num_nodes, num_taps = num_taps) rc_status = self .stub.rc_config(config_message) return message_to_dict(rc_status) def rc_run ( self, lock_id: str , reservoir_input: List [ int ], ): """ Runs a series of data through the reservoir and returns response from device based on current reservoir configuration. :param lock_id: a lock_id which has an active reserve on the device :param reservoir_input: a list of digitized values to input to the reservoir must be less than MAX_INPUT_SIZE :return: a dictionary with the folowing keys: - **status**: `int`- the status for the reservoir submission - **message**: `str`- a description of the status for the submission - **states**: `bytes`- response from reservoir as bytes. """ assert MAX_INPUT_SIZE>= len (reservoir_input), \ f"Input to reservoir must be of less than or equal to {MAX_INPUT_SIZE} was { len (reservoir_input)} " run_message = emucore_pb2.rc_run_message(lock_id=lock_id, input = reservoir_input,) rc_response = self .stub.rc_run(run_message) return message_to_dict(rc_response) def wait_for_lock ( self ): """ Waits for lock indefinitely calling :func:`acquire_lock` :return: a tuple of the following items: - **lock_id**: `str`- exclusive lock for device execution with a timeout - **start_queue_ts**: `int`- time in ns when began lock acquisition. - **end_queue_ts**: `int`- time in ns when lock was acquired. """ lock_id = "" start_queue_ts = time.time_ns() while lock_id == "" : lock_id = self .acquire_lock()[ "lock_id" ] if lock_id == "" : time.sleep( 1 ) end_queue_ts = time.time_ns() return lock_id, start_queue_ts, end_queue_ts def process_all_data ( self, lock_id: str , input_data: np.ndarray, num_nodes: int , density: float , feature_scaling: float , max_scale_val: float = None , weights: np.ndarray= None , seed_val_weights: int = 13 ): """ Run dataset through reservoir: 1. Get lock 2. Reset Reservoir 3. Calculate input coefficients 4. Run data through reservoir 5. Combine data from reservoir and reformat 6. Release lock :param lock_id: a UUID that currently has lock on the device :param input_data: data or series to process via reservoir :param num_taps: defines connectivity of neurons in the reservoir higher taps can lead to overfitting it is not recommended to not set num_taps significantly higher than num_nodes :param feature_scaling: after applying max abs scalar feature scaling factor applied :param max_scale_val: max absolute value used to scale data if provided :param seed_val_weights: seeds randomness for weigths to allow for reproducibility :note: if doing multiple runs without reset the max value mustn't exceed original data max value in order for results to register properly :return: a tuple of the following elements: - **reservoir_response**: `np.ndarray`- reservoir response represented as an array dimension of array will be nrows of input matrix by num nodes. - **max_scale_value**: `np.ndarray`- the scaling value that was applied to the input data before it was processed by the reservoir. - **weights**: `np.ndarray`- the weights that were used to apply the random mask to the data prior to being processed by the reservoir. """ n_rows, n_cols = input_data.shape input_packets, max_scale_val, weights = prep_input_data( input_data=input_data, num_nodes=num_nodes, density=density, feature_scaling=feature_scaling, max_scale_val=max_scale_val, weights=weights, seed_val_weights=seed_val_weights) try : reservoir_resp = np.array([]) for i in input_packets: packet_resp = self .rc_run( lock_id=lock_id, reservoir_input=i) packet_bytes = bytearray (np.array(packet_resp[ "states" ])) packet_arr = bytes_to_array(input_bytes=packet_bytes) reservoir_resp = np.concatenate((reservoir_resp, packet_arr),axis= 0 ) except Exception as err: print ( "ERROR OCCURRED" , err) raise RuntimeError( "Error while processing data" ) from err return np.array(reservoir_resp).reshape(n_rows,num_nodes), max_scale_val, weights