-
Notifications
You must be signed in to change notification settings - Fork 1
/
FILE_INFO
17 lines (11 loc) · 1.26 KB
/
FILE_INFO
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
FILE INFO:
domain_pretraining.sh: This script pretrains the user embeddings from the domain table and applies this learnt reduction to transform the train-test table provided.
The script takes in the reduction method, the reduction size along with the domain embeddings table and train-test embeddings table names.
task_traintest.sh: This script takes in the reduced train-test table embeddings name along with the outcomes and whether it is a regression or a classification flag.
This performs the training and testing.
results_formatter.py: This script takes in a txt file [output of DLATK after train and test] and creates a csv of the results.
It also takes in the metric to be searched for.
cmd_logs/pretraining_logs_%d.txt: This file contains the command logs to execute for the pretrainig of reduction on a particular domain, train-test table pairs.
Example - To execute RPCA of BERT embeddings of CLP18: eval "$(eval "$(sed -n '1,7p' < ~/NLP/ContextualEmbeddingDR3/pretraining_logs_18.txt)")"
cmd_logs/traintest_logs_%d.txt: Similar to the above, this contains the log for training and testing procedure of tasks.
Example - To collect results for a11_bsag_total and cntrl_gender in CLP18: eval "$(eval "$(sed -n '1,15p' < ~/NLP/ContextualEmbeddingDR3/traintest_logs_18.txt)")"