uda
uda copied to clipboard
google-research
Unsupervised Data Augmentation (UDA)
Model M is used to predict labels for unknown text, but I don't see how it was trained.
Hi, In the paper, it is written the UDA loss is performed by minimizing cross-entropy, but in the code, it is implemented in KL-divergence. Any further explanation about this? Many...
Hi, The link to download the cifar10 dataset is broken. Could you please update it? Thanks! ``` url_prefix=https://storage.googleapis.com/uda_model/image/proc_data_v2/${task_name} for sup_size in 4000 2000 1000 500 250; do wget $url_prefix/train-size_${sup_size}.tfrecord.0 done...
For IMDB dataset, how did you generate back_translation_file "/back_translation/imdb_back_trans/unsup_in/sample_0.9/para/para_0.txt"
While trying to run back_translate/download.sh, I get the following error: ``` > bash download.sh --2021-06-19 12:36:11-- https://storage.googleapis.com/uda_model/text/back_trans_checkpoints.zip Resolving storage.googleapis.com (storage.googleapis.com)... 172.217.8.16, 172.217.9.208, 172.217.12.240, ... Connecting to storage.googleapis.com (storage.googleapis.com)|172.217.8.16|:443... connected. HTTP...
I find it is hard to reproduce the result on my dataset. What are most important? 1, augmented data quality. 2, unsupervised data amount. 3, what hyperparameter? @qizhex Thank you...
Hello! I've found a performance issue in /image/data.py: `dataset = dataset.batch(per_core_bsz, drop_remainder=True)`[(here)](https://github.com/google-research/uda/blob/960684e363251772a5938451d4d2bc0f1da9e24b/image/data.py#L110) should be called before `dataset = dataset.map(parser, num_parallel_calls=32)`[(here)](https://github.com/google-research/uda/blob/960684e363251772a5938451d4d2bc0f1da9e24b/image/data.py#L109), which would make your program more efficient. Here is [the...
Hi, This script throws an error: scripts/download.sh The files are not available and wget is unable to reach them. I noticed that people have reported this issue already, but it...
some problem with 'wget https://storage.googleapis.com/uda_model/text/imdb_bert_ft.zip 'and 'wget https://storage.googleapis.com/uda_model/text/imdb_back_trans.zip' ,'wget https://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz' It doesn't work. How can I get the data from other sources? Thank you
