"""
Adapted from SentEval/examples/bow.py
https://github.com/facebookresearch/SentEval/blob/master/examples/bow.py
"""
import sys
import io
import numpy as np
import argparse
import logging
from tools.Blogger import Blogger
logger = Blogger()
# Set PATHs
PATH_TO_SENTEVAL = "./SentEval"
PATH_TO_DATA = "./SentEval/data"
# import SentEval
sys.path.insert(0, PATH_TO_SENTEVAL)
import senteval
# Create dictionary
def create_dictionary(sentences, threshold=0):
words = {}
for s in sentences:
for word in s:
words[word] = words.get(word, 0) + 1
if threshold > 0:
newwords = {}
from lib.ProcessEmbeddings import WordEmbeddings
from tools.Blogger import Blogger
import os
logger = Blogger()
CLASSIFICATION_TASKS = ["MR", "CR", "SUBJ", "MPQA", "SST5", "TREC", "MRPC"]
SIMILARITY_TASKS = [
"SICKRelatedness", "STS12", "STS13", "STS14", "STS15", "STS16"
]
if __name__ == "__main__":
WE = WordEmbeddings(vector_file="embeds/glove.6B.300d.txt")
WE.sparsify("ckpt/glove3000/ckpt-8000")
WE.subract_mean()
logger.status_update("Running SentEval tasks...")
WE.SentEval(
tasks=CLASSIFICATION_TASKS,
save_summary=True,
summary_file_name="glove_wta_3000.json",
)
"""
Run in Google Colab.
"""
from lib.ProcessEmbeddings import WordEmbeddings
from tools.Blogger import Blogger
from collections import defaultdict
logger = Blogger()
BINARY_CLASSIFICATION_TASKS = ["MR", "CR", "SUBJ", "MPQA", "SST2"]
MULTICLASS_CLASSIFICATION_TASKS = ["SST5", "TREC"]
CLASSIFICATION_TASKS = ["MR", "CR", "SUBJ", "MPQA", "SST2", "SST5", "TREC"]
SIMILARITY_TASKS = ['STS12', 'STS13', 'STS14']
ALL_TASKS = BINARY_CLASSIFICATION_TASKS + MULTICLASS_CLASSIFICATION_TASKS + SIMILARITY_TASKS
PRODUCTION_CONFIG = {
"usepytorch": True,
"kfold": 5,
"nhid": 0,
"optim": "rmsprop",
"batch_size": 128,
"tenacity": 3,
"epoch_size": 2
}
def glove(output_dir, dims, senteval_config):
summary_file_name = f"{output_dir}/glove_{dims}.json"
WE = WordEmbeddings(vector_file=f"embeds/glove.6B.{dims}d.txt")
# Default Glove
WE.evaluate(tasks=CLASSIFICATION_TASKS,
save_summary=True,
summary_file_name=summary_file_name,
overwrite_file=True,