Fix summarization postprocess backbone (#296)

* Fix summarization postprocess backbone

* Update CHANGELOG.md

* Update

CHANGELOG.md

@@ -31,6 +31,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 ### Fixed
 
 - Fix DataPipeline resolution in Task ([#212](https://github.com/PyTorchLightning/lightning-flash/pull/212))
+- Fixed a bug where the backbone used in summarization was not correctly passed to the postprocess ([#296](https://github.com/PyTorchLightning/lightning-flash/pull/296))
 
 
 ## [0.2.3] - 2021-04-17

flash/text/seq2seq/core/data.py

@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import os
+from dataclasses import dataclass
 from functools import partial
 from typing import Any, Callable, Dict, List, Optional, Union
 
@@ -24,6 +25,7 @@
 from flash.data.data_module import DataModule
 from flash.data.data_source import DataSource, DefaultDataSources
 from flash.data.process import Preprocess
+from flash.data.properties import ProcessState
 
 
 class Seq2SeqDataSource(DataSource):
@@ -158,6 +160,15 @@ def load_data(
         return [self._tokenize_fn(s) for s in data]
 
 
+@dataclass(unsafe_hash=True, frozen=True)
+class Seq2SeqBackboneState(ProcessState):
+    """The ``Seq2SeqBackboneState`` stores the backbone in use by the
+    :class:`~flash.text.seq2seq.core.data.Seq2SeqPreprocess`
+    """
+
+    backbone: str
+
+
 class Seq2SeqPreprocess(Preprocess):
 
     def __init__(
@@ -204,6 +215,8 @@ def __init__(
             default_data_source="sentences",
         )
 
+        self.set_state(Seq2SeqBackboneState(self.backbone))
+
     def get_state_dict(self) -> Dict[str, Any]:
         return {
             **self.transforms,

flash/text/seq2seq/summarization/data.py

@@ -16,19 +16,29 @@
 from transformers import AutoTokenizer
 
 from flash.data.process import Postprocess
-from flash.text.seq2seq.core.data import Seq2SeqData, Seq2SeqPreprocess
+from flash.text.seq2seq.core.data import Seq2SeqBackboneState, Seq2SeqData, Seq2SeqPreprocess
 
 
 class SummarizationPostprocess(Postprocess):
 
-    def __init__(
-        self,
-        backbone: str = "sshleifer/tiny-mbart",
-    ):
+    def __init__(self):
         super().__init__()
 
-        # TODO: Should share the backbone or tokenizer over state
-        self.tokenizer = AutoTokenizer.from_pretrained(backbone, use_fast=True)
+        self._backbone = None
+        self._tokenizer = None
+
+    @property
+    def backbone(self):
+        backbone_state = self.get_state(Seq2SeqBackboneState)
+        if backbone_state is not None:
+            return backbone_state.backbone
+
+    @property
+    def tokenizer(self):
+        if self.backbone is not None and self.backbone != self._backbone:
+            self._tokenizer = AutoTokenizer.from_pretrained(self.backbone, use_fast=True)
+            self._backbone = self.backbone
+        return self._tokenizer
 
     def uncollate(self, generated_tokens: Any) -> Any:
         pred_str = self.tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)

tests/text/summarization/test_data.py

@@ -64,7 +64,7 @@ def test_from_files(tmpdir):
         train_file=csv_path,
         val_file=csv_path,
         test_file=csv_path,
-        batch_size=1
+        batch_size=1,
     )
     batch = next(iter(dm.val_dataloader()))
     assert "labels" in batch
@@ -75,6 +75,25 @@ def test_from_files(tmpdir):
     assert "input_ids" in batch
 
 
+def test_postprocess_tokenizer(tmpdir):
+    """Tests that the tokenizer property in ``SummarizationPostprocess`` resolves correctly when a different backbone is
+    used.
+    """
+    backbone = "sshleifer/bart-tiny-random"
+    csv_path = csv_data(tmpdir)
+    dm = SummarizationData.from_csv(
+        "input",
+        "target",
+        backbone=backbone,
+        train_file=csv_path,
+        batch_size=1,
+    )
+    pipeline = dm.data_pipeline
+    pipeline.initialize()
+    assert pipeline._postprocess_pipeline.backbone == backbone
+    assert pipeline._postprocess_pipeline.tokenizer is not None
+
+
 @pytest.mark.skipif(os.name == "nt", reason="Huggingface timing out on Windows")
 def test_from_json(tmpdir):
     json_path = json_data(tmpdir)