wget https://s3-us-west-2.amazonaws.com/allennlp/models/elmo/vocab-2016-09-10.txt wget https://s3-us-west-2.amazonaws.com/allennlp/models/elmo/2x4096_512_2048cnn_2xhighway/elmo_2x4096_512_2048cnn_2xhighway_options.json wget https://s3-us-west-2.amazonaws.com/allennlp/models/elmo/2x4096_512_2048cnn_2xhighway/elmo_2x4096_512_2048cnn_2xhighway_weights.hdf5 mkdir models cd models mkdir curriculum curriculumline base corpuscurriculum cd curriculum mkdir 2 103 cd 2 mkdir baseline random unigram bigram trigram dep pos len cd ../103 mkdir baseline random unigram bigram trigram dep pos len cd ../../curriculumline mkdir 2 103 cd 2 mkdir baseline random unigram bigram trigram dep pos len cd ../103 mkdir baseline random unigram bigram trigram dep pos len cd ../../corpuscurriculum mkdir baseline-2 baseline-103 modified-2 modified-103 cd ../
cd .. #from https://blog.einstein.ai/the-wikitext-long-term-dependency-language-modeling-dataset/ wget https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-103-v1.zip wget https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-v1.zip unzip wikitext-2-v1.zip unzip wikitext-103-v1.zip cat wikitext-103/wiki.train.tokens wikitext-103/wiki.valid.tokens wikitext-103/wiki.test.tokens > wikitext-103/wikitext.tokens cat wikitext-2/wiki.train.tokens wikitext-2/wiki.valid.tokens wikitext-2/wiki.test.tokens > wikitext-2/wikitext.tokens rm wikitext-103-v1.zip rm wikitext-2-v1.zip
conda create -n ProgressiveTraining python=3.6 conda activate ProgressiveTraining conda install -c anaconda tensorflow-gpu=1.2.1 h5py spacy python -m spacy download en_core_web_lg python -m unittest discover tests/ export CUDA_VISIBLE_DEVICES=0,1,2
cd .. git clone --branch v1.3.1 --recursive https://github.com/nyu-mll/jiant.git jiant git checkout tags/v1.3.1 -b 1.3.1_master git submodule update --init --recursive conda env create -f environment.yml conda activate jiant python -m spacy download en python -m nltk.downloader perluniprops nonbreaking_prefixes punkt python scripts/download_glue_data.py --data_dir ../Data/ --tasks all python bin/make_vocab.py wikitext-103/wikitext.tokens wikitext-103/vocab.txt python bin/make_vocab.py wikitext-2/wikitext.tokens wikitext-2/vocab.txt python bin/make_sentence.py wikitext-103/wiki.train.tokens wikitext-103/wiki.train.tokens.sent python bin/make_sentence.py wikitext-103/wiki.valid.tokens wikitext-103/wiki.valid.tokens.sent python bin/make_sentence.py wikitext-103/wiki.test.tokens wikitext-103/wiki.test.tokens.sent python bin/make_sentence.py wikitext-2/wiki.train.tokens wikitext-2/wiki.train.tokens.sent python bin/make_sentence.py wikitext-2/wiki.test.tokens wikitext-2/wiki.test.tokens.sent python bin/make_sentence.py wikitext-2/wiki.valid.tokens wikitext-2/wiki.valid.tokens.sent
sh scripts/generate_competence_files.sh sh scripts/runcorpuscurriculum-wiki103.sh sh scripts/runcorpuscurriculum-wiki2.sh sh scripts/train_curicula-wiki-103.sh sh scripts/train_curicula-wiki-2.sh sh scripts/run_perplexity_eval_wikitext-103.sh sh scripts/run_perplexity_eval_wikitext-2.sh sh scripts/dump_all_weights.sh
source jiant_config.sh python main.py --config default.conf --overrides "elmo_weight_file_path=/home/spacemanidol/ProgressiveLanguageLearning/models/elmo_2x4096_512_2048cnn_2xhighway_weights.hdf5"
tail -n 1 Results/wikitext/bengiostylecurriculum/wikitext-103/corpus_curr_wiki103_validperplexity_epoch* | tr 'FINSIHED! AVERAGE PERPLEXITY = ' '' | tr '' 0
python consume_glue.py Results/glueresults/elmocompetencewiki2baseline.txt