Update fs.py

fs.py

@@ -6,7 +6,7 @@ import json
 import os
 import datasets
 from datasets import load_dataset
-from transformers import AutoTokenizer # TODO comment out when getting rid of __main__:
+from transformers import AutoTokenizer  # TODO comment out when getting rid of __main__:
 
 _FS_CITATION = """
 TBD
@@ -24,14 +24,14 @@ For SCROLLS, we use the ForeverDreaming (FD) subset, as it incorporates 88 diffe
 making it a more diverse alternative to the TV MegaSite (TMS) subset, which has only 10 shows. 
 Community-authored recaps for the ForeverDreaming transcripts were collected from English Wikipedia and TVMaze."""
 
-
 _GOV_REPORT_DESCRIPTION = """
 GovReport (Huang et al., 2021) is a summarization dataset of reports addressing various national policy issues published by the 
 Congressional Research Service and the U.S. Government Accountability Office, where each document is paired with a hand-written executive summary.
 The reports and their summaries are longer than their equivalents in other popular long-document summarization datasets; 
 for example, GovReport's documents are approximately 1.5 and 2.5 times longer than the documents in Arxiv and PubMed, respectively."""
 
-
+_ARXIV_DESCRIPTION = """
+"""
 
 _SUMM_SCREEN_CITATION = r"""
 @misc{chen2021summscreen,
@@ -62,11 +62,16 @@ _GOV_REPORT_CITATION = r"""
     abstract = "The quadratic computational and memory complexities of large Transformers have limited their scalability for long document summarization. In this paper, we propose Hepos, a novel efficient encoder-decoder attention with head-wise positional strides to effectively pinpoint salient information from the source. We further conduct a systematic study of existing efficient self-attentions. Combined with Hepos, we are able to process ten times more tokens than existing models that use full attentions. For evaluation, we present a new dataset, GovReport, with significantly longer documents and summaries. Results show that our models produce significantly higher ROUGE scores than competitive comparisons, including new state-of-the-art results on PubMed. Human evaluation also shows that our models generate more informative summaries with fewer unfaithful errors.",
 }"""
 
+_ARXIV_CITATION = r"""
+}"""
+
+SUMM_PROMPT = "Summary: "
+
 
 class FSConfig(datasets.BuilderConfig):
     """BuilderConfig for FS."""
 
-    def __init__(self, features, data_url, citation, url, max_source_length, tokenizer,prompt, **kwargs):
+    def __init__(self, features, data_url, citation, url, max_source_length, tokenizer, **kwargs):
         """BuilderConfig for FS.
         Args:
           features: `list[string]`, list of the features that will appear in the
@@ -86,27 +91,80 @@ class FSConfig(datasets.BuilderConfig):
         self.url = url
         self.max_source_length = max_source_length
         self.tokenizer = tokenizer
-        self.prompt = prompt
+        self.prompt = None
+        self.input_key = None
+        self.output_key = None
+        self.redundant_fields = []
+
+        self.train_file = "train.jsonl"
+        self.validation_file = "validation.jsonl"
+        self.test_file = "test.jsonl"
+
+    def remove_redundant_fields(self, example):
+        for field in self.redundant_fields:
+            del example[field]
+
+    def process_input(self, s):
+        return s.strip()
+
+    def process_output(self, s):
+        return s
+
+
+class ScrollsConfig(FSConfig):
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+        self.prompt = SUMM_PROMPT
+
+        self.train_file = "train.jsonl"
+        self.validation_file = "validation.jsonl"
+        self.test_file = "test.jsonl"
+
+        self.input_key = "input"
+        self.output_key = "output"
+        self.id_key = "pid"
+        self.redundant_fields = [self.input_key, self.output_key, "id"]
+
+
+class ArxivConfig(FSConfig):
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+        self.prompt = SUMM_PROMPT
+
+        self.train_file = "train.txt"
+        self.validation_file = "val.txt"
+        self.test_file = "test.txt"
+
+        self.input_key = "article_text"
+        self.output_key = "abstract_text"
+        self.id_key = "article_id"
+        self.redundant_fields = [self.input_key, self.output_key, self.id_key, 'labels', 'section_names', 'sections']
+
+    def process_input(self, s):
+        return ' '.join(s)
+
+    def process_output(self, s):
+        # TODO remove "<S>" and "</S>" ?
+        return self.process_input(s)
 
 
 class Fs(datasets.GeneratorBasedBuilder):
     """The SCROLLS benchmark."""
 
-    features = ["id", "pid", "input", "output"]
+    features = ["pid", "source", "target"]
     DEFAULT_WRITER_BATCH_SIZE = 1000  # because Narrative QA is a rather large dataset
     BUILDER_CONFIGS = [
-        FSConfig(
-            name="summ_screen_fd",
+        ScrollsConfig(
+            name="summ_screen_fd_debug",
             description=_SUMM_SCREEN_DESCRIPTION,
             features=features,
-            data_url="https://huggingface.co/datasets/tau/fs/resolve/main/data/summ_screen_fd.zip",
+            data_url="https://huggingface.co/datasets/tau/fs/resolve/main/data/summ_screen_fd_debug.zip",
             citation=_SUMM_SCREEN_CITATION,
             url="https://github.com/mingdachen/SummScreen",
             max_source_length=None,
             tokenizer=None,
-            prompt=None
         ),
-        FSConfig(
+        ScrollsConfig(
             name="gov_report",
             description=_GOV_REPORT_CITATION,
             features=features,
@@ -115,7 +173,16 @@ class Fs(datasets.GeneratorBasedBuilder):
             url="https://gov-report-data.github.io/",
             max_source_length=None,
             tokenizer=None,
-            prompt=None
+        ),
+        ArxivConfig(
+            name="arxiv_debug",
+            description=_ARXIV_CITATION,
+            features=features,
+            data_url="https://huggingface.co/datasets/tau/fs/resolve/main/data/arxiv_debug.zip",
+            citation=_ARXIV_DESCRIPTION,
+            url="https://github.com/armancohan/long-summarization",
+            max_source_length=None,
+            tokenizer=None,
         ),
     ]
 
@@ -141,30 +208,28 @@ class Fs(datasets.GeneratorBasedBuilder):
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={
-                    "data_file": os.path.join(dl_dir, "train.jsonl"),
+                    "data_file": os.path.join(dl_dir, self.config.train_file),
                     "split": datasets.Split.TRAIN,
                     "max_source_length": self.config.max_source_length,
-                    "prompt": self.config.prompt,
                     "tokenizer": self.config.tokenizer,
                 },
             ),
             datasets.SplitGenerator(
                 name=datasets.Split.VALIDATION,
                 gen_kwargs={
-                    "data_file": os.path.join(dl_dir, "validation.jsonl"),
+                    "data_file": os.path.join(dl_dir, self.config.validation_file),
                     "split": datasets.Split.VALIDATION,
                     "max_source_length": self.config.max_source_length,
-                    "prompt": self.config.prompt,
                     "tokenizer": self.config.tokenizer,
                 },
             ),
             datasets.SplitGenerator(
                 name=datasets.Split.TEST,
                 gen_kwargs={
-                    "data_file": os.path.join(dl_dir, "test.jsonl") if data_files is None else data_files["test"],
+                    "data_file": os.path.join(dl_dir, self.config.test_file) if data_files is None else data_files[
+                        "test"],
                     "split": datasets.Split.TEST,
                     "max_source_length": self.config.max_source_length,
-                    "prompt": self.config.prompt,
                     "tokenizer": self.config.tokenizer,
                 },
             ),
@@ -174,7 +239,8 @@ class Fs(datasets.GeneratorBasedBuilder):
         with open(data_file, encoding="utf-8") as f:
             for line in f:
                 row = json.loads(line)
-                prefix = row["source"].strip()
+
+                prefix = self.config.process_input(row[self.config.input_key])
                 suffix = "\n" + self.config.prompt
                 encoded_input = tokenizer.encode(prefix + suffix)
 
@@ -188,9 +254,12 @@ class Fs(datasets.GeneratorBasedBuilder):
                     prefix = tokenizer.decode(tokenized_prefix, skip_special_tokens=False).strip()
                     encoded_input = tokenizer.encode(prefix + suffix)
 
-                row["input"] = prefix + suffix
-                row["output"] = row["target"]
+                row["pid"] = row[self.config.id_key]
+                row["source"] = prefix + suffix
+                target = row[self.config.output_key]
+                row["target"] = self.config.process_output(target)
 
+                self.config.remove_redundant_fields(row)
                 yield row["pid"], row
 
 
@@ -200,7 +269,11 @@ def _get_task_name_from_data_url(data_url):
 
 if __name__ == '__main__':
     tokenizer = AutoTokenizer.from_pretrained("t5-base")
-    dataset = load_dataset("tau/fs",name="summ_screen_fd", max_source_length=512, tokenizer=tokenizer, prompt="Summary:")
+    # dataset = load_dataset("tau/fs",name="summ_screen_fd", max_source_length=512, tokenizer=tokenizer, prompt="Summary:")
+    ssfd_debug = load_dataset("/Users/uri/Projects/fs/fs.py", name="summ_screen_fd_debug", max_source_length=512,
+                              tokenizer=tokenizer)
+    arxiv_debug = load_dataset("/Users/uri/Projects/fs/fs.py", name="arxiv_debug", max_source_length=512,
+                               tokenizer=tokenizer, prompt="Summarize the above:")
     x = 5
 #    builder = Scrolls("scrolls", "summ_screen_fd")
 #    builder.download_and_prepare()