#161 update query size

README.md

@@ -96,7 +96,7 @@ model = ModelWrapper(model, your_criterion)
 active_loop = ActiveLearningLoop(dataset,
                                  get_probabilities=model.predict_on_dataset,
                                  heuristic=heuristics.BALD(shuffle_prop=0.1),
-                                 ndata_to_label=NDATA_TO_LABEL)
+                                 query_size=NDATA_TO_LABEL)
 for al_step in range(N_ALSTEP):
     model.train_on_dataset(dataset, optimizer, BATCH_SIZE, use_cuda=use_cuda)
     if not active_loop.step():

baal/active/active_loop.py

@@ -1,14 +1,17 @@
 import os
 import pickle
 import types
+import warnings
 from typing import Callable
 
 import numpy as np
+import structlog
 import torch.utils.data as torchdata
 
 from . import heuristics
 from .dataset import ActiveLearningDataset
 
+log = structlog.get_logger(__name__)
 pjoin = os.path.join
 
 
@@ -20,7 +23,7 @@ class ActiveLearningLoop:
         get_probabilities (Function): Dataset -> **kwargs ->
                                         ndarray [n_samples, n_outputs, n_iterations].
         heuristic (Heuristic): Heuristic from baal.active.heuristics.
-        ndata_to_label (int): Number of sample to label per step.
+        query_size (int): Number of sample to label per step.
         max_sample (int): Limit the number of sample used (-1 is no limit).
         uncertainty_folder (Optional[str]): If provided, will store uncertainties on disk.
         **kwargs: Parameters forwarded to `get_probabilities`.
@@ -31,12 +34,18 @@ def __init__(
         dataset: ActiveLearningDataset,
         get_probabilities: Callable,
         heuristic: heuristics.AbstractHeuristic = heuristics.Random(),
-        ndata_to_label: int = 1,
+        query_size: int = 1,
         max_sample=-1,
         uncertainty_folder=None,
+        ndata_to_label=None,
         **kwargs,
     ) -> None:
-        self.ndata_to_label = ndata_to_label
+        if ndata_to_label is not None:
+            warnings.warn(
+                "`ndata_to_label` is deprecated, please use `query_size`.", DeprecationWarning
+            )
+            query_size = ndata_to_label
+        self.query_size = query_size
         self.get_probabilities = get_probabilities
         self.heuristic = heuristic
         self.dataset = dataset
@@ -88,6 +97,6 @@ def step(self, pool=None) -> bool:
                         open(pjoin(self.uncertainty_folder, uncertainty_name), "wb"),
                     )
                 if len(to_label) > 0:
-                    self.dataset.label(to_label[: self.ndata_to_label])
+                    self.dataset.label(to_label[: self.query_size])
                     return True
         return False

baal/utils/pytorch_lightning.py

@@ -1,5 +1,6 @@
 import sys
 import types
+import warnings
 from collections.abc import Sequence
 from typing import Dict, Any, Optional
 
@@ -19,6 +20,12 @@
 
 log = structlog.get_logger("PL testing")
 
+warnings.warn(
+    "baal.utils.pytorch_lightning is deprecated. BaaL is now integrated into Lightning Flash!"
+    " Please see experiments/pytorch_lightning/lightning_flash_example.py for a new tutorial!",
+    DeprecationWarning,
+)
+
 
 class BaaLDataModule(LightningDataModule):
     def __init__(self, active_dataset: ActiveLearningDataset, batch_size=1, **kwargs):
@@ -99,7 +106,7 @@ class BaalTrainer(Trainer):
     Args:
         dataset (ActiveLearningDataset): Dataset with some sample already labelled.
         heuristic (Heuristic): Heuristic from baal.active.heuristics.
-        ndata_to_label (int): Number of sample to label per step.
+        query_size (int): Number of sample to label per step.
         max_sample (int): Limit the number of sample used (-1 is no limit).
         **kwargs: Parameters forwarded to `get_probabilities`
             and to pytorch_ligthning Trainer.__init__
@@ -109,12 +116,12 @@ def __init__(
         self,
         dataset: ActiveLearningDataset,
         heuristic: heuristics.AbstractHeuristic = heuristics.Random(),
-        ndata_to_label: int = 1,
+        query_size: int = 1,
         **kwargs
     ) -> None:
 
         super().__init__(**kwargs)
-        self.ndata_to_label = ndata_to_label
+        self.query_size = query_size
         self.heuristic = heuristic
         self.dataset = dataset
         self.kwargs = kwargs
@@ -194,6 +201,6 @@ def step(self, model=None, datamodule: Optional[BaaLDataModule] = None) -> bool:
             if probs is not None and (isinstance(probs, types.GeneratorType) or len(probs) > 0):
                 to_label = self.heuristic(probs)
                 if len(to_label) > 0:
-                    self.dataset.label(to_label[: self.ndata_to_label])
+                    self.dataset.label(to_label[: self.query_size])
                     return True
         return False

experiments/mlp_mcdropout.py

@@ -44,7 +44,7 @@
     dataset=al_dataset,
     get_probabilities=wrapper.predict_on_dataset,
     heuristic=bald,
-    ndata_to_label=100,  # We will label 100 examples per step.
+    query_size=100,  # We will label 100 examples per step.
     # KWARGS for predict_on_dataset
     iterations=20,  # 20 sampling for MC-Dropout
     batch_size=32,

experiments/mlp_regression_mcdropout.py

@@ -84,7 +84,7 @@ def __getitem__(self, item):
     dataset=al_dataset,
     get_probabilities=wrapper.predict_on_dataset,
     heuristic=variance,
-    ndata_to_label=250,  # We will label 20 examples per step.
+    query_size=250,  # We will label 20 examples per step.
     # KWARGS for predict_on_dataset
     iterations=20,  # 20 sampling for MC-Dropout
     batch_size=16,

experiments/nlp_bert_mcdropout.py

@@ -30,7 +30,7 @@ def parse_args():
     parser.add_argument("--batch_size", default=32, type=int)
     parser.add_argument("--initial_pool", default=1000, type=int)
     parser.add_argument("--model", default="bert-base-uncased", type=str)
-    parser.add_argument("--n_data_to_label", default=100, type=int)
+    parser.add_argument("--query_size", default=100, type=int)
     parser.add_argument("--heuristic", default="bald", type=str)
     parser.add_argument("--iterations", default=20, type=int)
     parser.add_argument("--shuffle_prop", default=0.05, type=float)
@@ -112,7 +112,7 @@ def main():
         active_set,
         model.predict_on_dataset,
         heuristic,
-        hyperparams.get("n_data_to_label", 1),
+        hyperparams.get("query_size", 1),
         iterations=hyperparams["iterations"],
     )
 

experiments/pytorch_lightning/active_image_classification.py

@@ -153,7 +153,7 @@ def main():
         max_epochs=args.training_duration,
         logger=logger,
         heuristic=heuristic,
-        ndata_to_label=args.query_size,
+        query_size=args.query_size,
     )
 
     AL_STEPS = 100

experiments/segmentation/unet_mcdropout_pascal.py

@@ -39,7 +39,7 @@ def parse_args():
     parser.add_argument("--al_step", default=200, type=int)
     parser.add_argument("--batch_size", default=8, type=int)
     parser.add_argument("--initial_pool", default=40, type=int)
-    parser.add_argument("--n_data_to_label", default=20, type=int)
+    parser.add_argument("--query_size", default=20, type=int)
     parser.add_argument("--lr", default=0.001)
     parser.add_argument("--heuristic", default="random", type=str)
     parser.add_argument("--reduce", default="sum", type=str)
@@ -134,7 +134,7 @@ def main():
         active_set,
         model.predict_on_dataset_generator,
         heuristic=heuristic,
-        ndata_to_label=hyperparams["n_data_to_label"],
+        query_size=hyperparams["query_size"],
         # Instead of predicting on the entire pool, only a subset is used
         max_sample=1000,
         batch_size=batch_size,

experiments/ssl_experiments/pimodel_mcdropout_cifar10.py

@@ -100,7 +100,7 @@ def add_model_specific_args(parent_parser):
         callbacks=[ResetCallback(copy.deepcopy(model.state_dict()))],
         dataset=active_set,
         heuristic=heuristic,
-        ndata_to_label=params.query_size,
+        query_size=params.query_size,
     )
 
     AL_STEPS = 2000

experiments/vgg_mcdropout_cifar10.py

@@ -12,8 +12,8 @@
 from torchvision.transforms import transforms
 from tqdm import tqdm
 
-from baal import get_heuristic, ActiveLearningDataset
-from baal import ActiveLearningLoop
+from baal.active import get_heuristic, ActiveLearningDataset
+from baal.active.active_loop import ActiveLearningLoop
 from baal.bayesian.dropout import patch_module
 from baal import ModelWrapper
 
@@ -27,7 +27,7 @@ def parse_args():
     parser.add_argument("--epoch", default=100, type=int)
     parser.add_argument("--batch_size", default=32, type=int)
     parser.add_argument("--initial_pool", default=1000, type=int)
-    parser.add_argument("--n_data_to_label", default=100, type=int)
+    parser.add_argument("--query_size", default=100, type=int)
     parser.add_argument("--lr", default=0.001)
     parser.add_argument("--heuristic", default="bald", type=str)
     parser.add_argument("--iterations", default=20, type=int)
@@ -107,7 +107,7 @@ def main():
         active_set,
         model.predict_on_dataset,
         heuristic,
-        hyperparams.get("n_data_to_label", 1),
+        hyperparams.get("query_size", 1),
         batch_size=10,
         iterations=hyperparams["iterations"],
         use_cuda=use_cuda,

notebooks/compatibility/pytorch_lightning.ipynb

@@ -246,7 +246,7 @@
    "source": [
     "trainer = BaalTrainer(dataset=data_module.active_dataset,\n",
     "                      heuristic=heuristic,\n",
-    "                      ndata_to_label=hparams.query_size,\n",
+    "                      query_size=hparams.query_size,\n",
     "                      max_epochs=10, default_root_dir=hparams.data_root,\n",
     "                      gpus=hparams.gpus,\n",
     "                      callbacks=[ResetCallback(copy.deepcopy(model.state_dict()))])"

notebooks/compatibility/sklearn_tutorial.ipynb

@@ -261,9 +261,9 @@
     "  print(\"Test Accuracy  :: \", accuracy_score(test_y, predictions))\n",
     "  probs = predict(dataset.pool, clf)\n",
     "  to_label = heuristic(probs)\n",
-    "  ndata_to_label = 10\n",
+    "  query_size = 10\n",
     "  if len(to_label) > 0:\n",
-    "      dataset.label(to_label[: ndata_to_label])\n",
+    "      dataset.label(to_label[: query_size])\n",
     "  else:\n",
     "    break"
    ],