[README] add clip compressor.

README.md

@@ -197,22 +197,25 @@ bin/clip/main_small.sh
 
 The following table comes directly from the results which are automatically saved at `results/exp_clip_bottleneck_linear_eval/**/datapred_*/**/results_predictor.csv`. It shows the result of compression from our CLIP compressor on many datasets.
 
-|               | Cars196 | STL10 | Caltech101 | Food101 | PCam | Pets37 | CIFAR10 |
-|---------------|:-------:|:-----:|:----------:|:-------:|:----:|:------:|:-------:|
-| Rate [bits]   |   1468  |  1344 |    1341    |   1269  | 1491 |  1211  |   1408  |
-| Test Acc. [%] |   79.9  |  98.7 |    93.7    |   83.6  | 81.1 |  88.3  |   94.8  |
+|               | Cars196 | STL10 | Caltech101 | Food101 | PCam | Pets37 | CIFAR10 | CIFAR100 |
+|---------------|:-------:|:-----:|:----------:|:-------:|:----:|:------:|:-------:|:--------:|
+| Rate [bits]   |   1471  |  1342 |    1340    |   1266  | 1491 |  1209  |   1407  |   1413   |
+| Test Acc. [%] |   80.3  |  98.5 |    93.3    |   83.8  | 81.1 |  88.8  |   94.6  |   79.0   |
 
-Note: ImageNet is too large for training a SVM using SKlearn. You need to run MLP evaluation with `bin/clip/clip_bottleneck_mlp_eval`.
+Note: ImageNet is too large for training a SVM using SKlearn. You need to run MLP evaluation with `bin/clip/clip_bottleneck_mlp_eval`. Also you have to download ImageNet manually.
 
 
 ## Cite
-```
+
+You can read the full paper [here](https://arxiv.org/abs/2106.10800). Please cite our paper if you use our model:
+
+```bibtex
 @inproceedings{
-dubois2021lossy,
-title={Lossy Compression for Lossless Prediction},
-author={Yann Dubois and Benjamin Bloem-Reddy and Karen Ullrich and Chris J. Maddison},
-booktitle={Neural Compression: From Information Theory to Applications -- Workshop @ ICLR 2021},
-year={2021},
-url={https://arxiv.org/abs/2106.10800}
+    dubois2021lossy,
+    title={Lossy Compression for Lossless Prediction},
+    author={Yann Dubois and Benjamin Bloem-Reddy and Karen Ullrich and Chris J. Maddison},
+    booktitle={Neural Compression: From Information Theory to Applications -- Workshop @ ICLR 2021},
+    year={2021},
+    url={https://arxiv.org/abs/2106.10800}
 }
-```
+```

bin/clip/clip_bottleneck_linear_eval.sh

@@ -31,6 +31,7 @@ kwargs_multi="
 data@data_pred=stl10,cars196,caltech101,food101,pcam,pets37,cifar10,cifar100
 " 
 
+
 if [ "$is_plot_only" = false ] ; then
   for beta in   "5e-02" 
   do

lossyless/classical_compressors.py

@@ -1,15 +1,15 @@
 import logging
 
 import PIL
-
-import pytorch_lightning as pl
 import torch
 import torch.nn as nn
-from compressai.utils.bench import codecs
-from pytorch_lightning.core.decorators import auto_move_data
 from torch.nn import functional as F
 from torchvision.transforms import ToPILImage, ToTensor
 
+import pytorch_lightning as pl
+from compressai.utils.bench import codecs
+from pytorch_lightning.core.decorators import auto_move_data
+
 from .helpers import UnNormalizer, dict_mean, rename_keys_
 
 __all__ = ["ClassicalCompressor"]
@@ -165,7 +165,6 @@ def __init__(self, hparams):
 
         self.stage = self.hparams.stage  # allow changing to stages
 
-    @auto_move_data  # move data on correct device for inference
     def forward(self, x, is_return_out=False, **kwargs):
         """Represents the data `x`.
 

lossyless/learnable_compressors.py

@@ -10,7 +10,8 @@
 from .architectures import get_Architecture
 from .distortions import get_distortion_estimator
 from .distributions import CondDist, Deterministic
-from .helpers import BASE_LOG, Annealer, OrderedSet, Timer, append_optimizer_scheduler_
+from .helpers import (BASE_LOG, Annealer, OrderedSet, Timer,
+                      append_optimizer_scheduler_)
 from .predictors import OnlineEvaluator
 from .rates import get_rate_estimator
 
@@ -112,7 +113,6 @@ def predict_step(self, batch, batch_idx, dataloader_idx=None):
     def predict(self, *args, **kwargs):  # TODO remove in newer version of lightning
         return self.predict_step(*args, **kwargs)
 
-    @auto_move_data  # move data on correct device for inference
     def forward(self, x, is_compress=False, is_features=None):
         """Represents the data `x`.
 

lossyless/predictors.py

@@ -68,7 +68,6 @@ def __init__(self, hparams, featurizer=None):
 
         self.stage = self.hparams.stage
 
-    @auto_move_data  # move data on correct device for inference
     def forward(self, x, is_logits=True, is_return_logs=False):
         """Perform prediction for `x`.
 

utils/Z_linear_eval.py

@@ -78,8 +78,10 @@ def main(cfg):
         n_iter=cfg.clip_linear.n_predictors,
     )
 
+    logger.info(f"Training classifier.")
     clf.fit(Z, Y)
 
+    logger.info(f"Evaluating classifier.")
     Y_pred = clf.predict(Z_test)
 
     bacc = balanced_accuracy_score(Y_test, Y_pred)
@@ -91,6 +93,9 @@ def main(cfg):
     analyser.results["test/pred/balanced_err"] = 1 - bacc
 
     # save results
+    path = Path(analyser.cfgs[stage].paths.results)
+    path.mkdir(parents=True, exist_ok=True)
+    results_file = path / RESULTS_FILE.format(stage=stage)
     test_res_rep = replace_keys(analyser.results, "test/", "")
     tosave = dict(test=test_res_rep)
     results = pd.DataFrame.from_dict(tosave)