liulu112601/urt

ICLR 2021: A Universal Representation Transformer Layer for Few-Shot Image Classification

urt's Issues

'ImageDecoder' object has no attribute 'out_type'

I am trying to get the pretrained features myself using the pre-extract-feature.py. However, got this bug. It seems a bug due to the version of the tensorflow. I use the tensorflow 1.14 but still got this error.

why is it that using the pre-training model you provided, without any changes in the code, the test results vary greatly, even up to 10 point fluctuations？

Hi, Thank you for sharing your code. But why is it that using the pre-training model you provided, without any changes in the code, the test results vary greatly, even up to 10 point fluctuations？May I ask how the test results provided in your paper can be determined as the final result when the performance fluctuates so much？ Looking forward to your reply.

No values supplied by Gin or caller for arguments during pre-extract-feature.sh execution

Hi there!

I am trying to run the feature dump script (it is not easy for me to download such large folders from google drive in the cluster I run my experiments for now), but I am getting this missing-argument error message below.

Could you guide me on which values should I use for ignore_hierarchy_probabilityand simclr_episode_fraction. Oh, and can you show me as well how I insert them?

By the way, I loved your paper.

Thanks in advance for your time :)

2021-09-15 12:14:12.036526: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
[2021-09-15 04:14:23 PM] --- args ---
  [save_dir  ] : features_cache
  [model.backbone] : resnet18
  [model.classifier] : cosine
  [train.max_iter] : 10000
  [eval.max_iter] : 600
[2021-09-15 04:14:23 PM] --- args ---
Traceback (most recent call last):
  File "exps/pre-extract-feature.py", line 144, in <module>
    main(xargs)
  File "exps/pre-extract-feature.py", line 91, in main
    train_loader_lst  = [MetaDatasetEpisodeReader('train', [d], [d], all_test_datasets) for d in extractor_domains]
  File "exps/pre-extract-feature.py", line 91, in <listcomp>
    train_loader_lst  = [MetaDatasetEpisodeReader('train', [d], [d], all_test_datasets) for d in extractor_domains]
  File "/dccstor/kcsys/brenow/utils/URT/fast-exps/lib/data/meta_dataset_reader.py", line 110, in __init__
    train_episode_desscription = config.EpisodeDescriptionConfig(None, None, None)
  File "/u/brenow1/miniconda3/envs/neurips_2021/lib/python3.8/site-packages/gin/config.py", line 1069, in gin_wrapper
    utils.augment_exception_message_and_reraise(e, err_str)
  File "/u/brenow1/miniconda3/envs/neurips_2021/lib/python3.8/site-packages/gin/utils.py", line 41, in augment_exception_message_and_reraise
    raise proxy.with_traceback(exception.__traceback__) from None
  File "/u/brenow1/miniconda3/envs/neurips_2021/lib/python3.8/site-packages/gin/config.py", line 1046, in gin_wrapper
    return fn(*new_args, **new_kwargs)
TypeError: __init__() missing 2 required positional arguments: 'ignore_hierarchy_probability' and 'simclr_episode_fraction'
  No values supplied by Gin or caller for arguments: ['ignore_hierarchy_probability', 'simclr_episode_fraction']
  Gin had values bound for: ['ignore_bilevel_ontology', 'ignore_dag_ontology', 'max_log_weight', 'max_num_query', 'max_support_set_size', 'max_support_size_contrib_per_class', 'max_ways_upper_bound', 'min_log_weight', 'min_ways']
  Caller supplied values for: ['num_query', 'num_support', 'num_ways', 'self']
  In call to configurable 'EpisodeDescriptionConfig' (<class 'meta_dataset.data.config.EpisodeDescriptionConfig'>)