The goal of parscanlogreader is to read and process raw log files from Scikit-learn’s RandomizedSearchCV.
The development version can be installed from GitHub with:
devtools::install_github("kreh-team/parscanlogreader")
This is a basic example which shows you how the data pipeline works:
library(parscanlogreader)
src_file <- "logs/cnn-gru-scan.log"
log_data_raw <- src_file %>%
read_raw_log() %>%
clean_log_data()
log_data <- log_data_raw %>%
summarise_log_data()
Note that the functions are able to automatically parse the parameters
params_list
, numeric_params
, num_folds
, and num_models
from the
raw log files.
If you want, you can manually set them yourself, as shown in the example below:
src_params_list <- c(
"optimizers", "opt_recurrent_regs", "opt_kernel_regs", "opt_go_backwards",
"opt_dropout_recurrent", "opt_dropout", "maxpool_size", "kernel_size",
"gru_hidden_units", "filter_conv", "epochs", "batch_size", "activation_conv"
)
src_numeric_params <- c(
"opt_dropout_recurrent", "opt_dropout", "maxpool_size", "kernel_size",
"gru_hidden_units", "filter_conv", "epochs", "batch_size"
)
log_data_raw <- src_file %>%
read_raw_log(
params_list = src_params_list,
numeric_params = src_numeric_params
) %>%
clean_log_data()
log_data <- log_data_raw %>%
tidyr::drop_na() %>%
summarise_log_data(
params_list = src_params_list,
num_folds = 5,
num_models = 50
)