tests/testthat/test-vwsetup.R

In ivan-pavlov/rvwgsoc: R Interface for Vowpal Wabbit

context("vwsetup")
library(rvw)

test_model <- list(params = list(algorithm = "sgd",
                                 general_params = list(random_seed=0,
                                                       ring_size=NA_real_,
                                                       holdout_off=FALSE,
                                                       holdout_period=10,
                                                       holdout_after=0,
                                                       early_terminate=3,
                                                       loss_function=NA_character_,
                                                       link=NA_character_,
                                                       quantile_tau=0.5),
                                 feature_params = list(bit_precision=18,
                                                       quadratic=NA_character_,
                                                       cubic=NA_character_,
                                                       interactions=NA_character_,
                                                       permutations=FALSE,
                                                       leave_duplicate_interactions=FALSE,
                                                       noconstant=FALSE, 
                                                       feature_limit=NA_character_,
                                                       ngram=NA_character_,
                                                       skips=NA_character_,
                                                       hash=NA_character_,
                                                       affix=NA_character_,
                                                       spelling=NA_character_,
                                                       interact=NA_character_),
                                 optimization_params = list(adaptive=TRUE,
                                                            normalized=TRUE,
                                                            invariant=TRUE,
                                                            adax=FALSE,
                                                            sparse_l2=0,
                                                            l1_state=0,
                                                            l2_state=1,
                                                            learning_rate=0.5,
                                                            initial_pass_length=NA_real_,
                                                            l1=0,
                                                            l2=0,
                                                            no_bias_regularization=NA_character_,
                                                            feature_mask=NA_character_,
                                                            decay_learning_rate=1,
                                                            initial_t=0,
                                                            power_t=0.5,
                                                            initial_weight=0,
                                                            random_weights="off",
                                                            normal_weights="off",
                                                            truncated_normal_weights="off",
                                                            sparse_weights=FALSE,
                                                            input_feature_regularizer=NA_character_),
                                 options = list()
                                 
),
dir = "../my_tmp",
model = "mdl.vw",
params_str = paste0(""),
is_cl = FALSE,
data_md5sum = list(train = "",
                   test = ""),
train_file = "",
eval = list(
    train=list(
        num_examples = NA_real_,
        weighted_example_sum = NA_real_,
        weighted_label_sum = NA_real_,
        avg_loss = NA_real_,
        avg_multiclass_log_loss = NA_real_,
        best_const = NA_real_,
        best_const_loss = NA_real_,
        total_feature = NA_real_
    ),
    test=list(
        num_examples = NA_real_,
        weighted_example_sum = NA_real_,
        weighted_label_sum = NA_real_,
        avg_loss = NA_real_,
        avg_multiclass_log_loss = NA_real_,
        best_const = NA_real_,
        best_const_loss = NA_real_,
        total_feature = NA_real_
    )
),
parser_opts=NA
)
class(test_model) <- "vw"

test_that("vwsetup correctly setup model with different learning modes", {
    # Empty setup
    expect_equal(vwsetup(dir = "../my_tmp/", model = "mdl.vw"), test_model)
    
    # CMD line setup
    cmd_test_model <- test_model
    cmd_test_model$params_str <- "-b 25 --link glm"
    cmd_test_model$is_cl <- TRUE
    cmd_test_model$params$general_params <- list()
    cmd_test_model$params$feature_params <- list()
    cmd_test_model$params$optimization_params <- list()
    cmd_test_model$params$algorithm <- NA
    expect_equal(vwsetup(dir = "../my_tmp/", model = "mdl.vw", params_str = "-b 25 --link glm"),
                 cmd_test_model)
    
    # Reference test model for nn mode
    nn_test_model <- test_model
    nn_test_model$params$options = list(nn = list(num_hidden=3,
                                                  inpass=FALSE,
                                                  multitask=FALSE,
                                                  dropout=FALSE,
                                                  meanfield=FALSE))
    nn_test_model$params_str = paste0("--nn 3")
    expect_equal(
        vwsetup(dir = "../my_tmp/", model = "mdl.vw", option = "nn", num_hidden=3),
        nn_test_model
    )
    
    # Reference test model for lda mode
    lda_test_model <- test_model
    lda_test_model$params$options = list(lda = list(num_topics=5,
                                                    lda_alpha=0.100000001,
                                                    lda_rho=0.100000001,
                                                    lda_D=10000,
                                                    lda_epsilon=0.00100000005,
                                                    math_mode=NA_character_,
                                                    minibatch=1,
                                                    metrics=0))
    lda_test_model$params_str = paste0("--lda 5")
    
    expect_equal(
        vwsetup(dir = "../my_tmp/", model = "mdl.vw", option = "lda", num_topics=5),
        lda_test_model
    )
    
    # Reference test model with custom parameters
    custom_test_model <- test_model
    custom_test_model$params$optimization_params$adaptive = FALSE
    custom_test_model$params$options = list(binary = list(binary=TRUE))
    custom_test_model$params_str <- paste0("--binary")
    # Package vwmodel setup
    test_vwmodel <- vwsetup(
        dir = "../my_tmp/",
        model = "mdl.vw",
        optimization_params = list(adaptive=FALSE),
        option = "binary"
    )
    expect_equal(test_vwmodel, custom_test_model)
    
    # Reference test model with Experience Replay
    replay_test_model <- test_model
    replay_test_model$params$options = list(replay = list(level="m",
                                                          buffer=200,
                                                          count=1))
    replay_test_model$params_str <- paste0("--replay_m 200 --replay_m_count 1")
    # Package vwmodel setup
    test_vwmodel <- vwsetup(
        dir = "../my_tmp/",
        model = "mdl.vw",
        option = "replay",
        level="m",
        buffer=200
    )
    expect_equal(test_vwmodel, replay_test_model)
    
    # Reference test model with Contextual Bandit Exploration with Action Dependent Features
    cb_explore_test_model <- test_model
    cb_explore_test_model$params$options = list(cb_explore = list(num_actions=0,
                                                                  explore_type="bag",
                                                                  explore_arg=10,
                                                                  psi=1,
                                                                  nounif=FALSE,
                                                                  mellowness=0.1,
                                                                  greedify=FALSE,
                                                                  lambda=-1,
                                                                  cb_min_cost=0,
                                                                  cb_max_cost=1,
                                                                  first_only=FALSE))
    cb_explore_test_model$params_str <- paste0("--cb_explore_adf --bag 10")
    # Package vwmodel setup
    test_vwmodel <- vwsetup(
        dir = "../my_tmp/",
        model = "mdl.vw",
        option = "cb_explore",
        num_actions=0,
        explore_type="bag",
        explore_arg=10
    )
    expect_equal(test_vwmodel, cb_explore_test_model)
})