first attempt to prompt and truncate the suffix of the input

fs.py

@@ -168,6 +168,25 @@ class Fs(datasets.GeneratorBasedBuilder):
         with open(data_file, encoding="utf-8") as f:
             for line in f:
                 row = json.loads(line)
+
+                prefix, addition = row["source"].strip(), "Summary:"
+                input_length = len(tokenizer.encode(prefix + "\n" + addition))
+
+                n_truncations = 0
+                while input_length > max_source_length:
+                    overflow = len(tokenizer.encode(prefix + "\n" + addition)) - max_source_length
+                    tokenized_prefix = tokenizer.encode(prefix, add_special_tokens=False)
+                    if overflow > 0:
+                        tokenized_prefix = tokenized_prefix[:-overflow]
+                    n_truncations += 1
+                    prefix = tokenizer.decode(tokenized_prefix, skip_special_tokens=False).strip()
+                    input_length = len(tokenizer.encode(prefix + "\n" + addition))
+
+                valid_input = prefix + "\n" + addition
+
+                row["input"] = valid_input
+                row["output"] = row["target"]
+
                 yield row["pid"], row