[breaking change] refactor data pipeline (#6901)

* refactor data

* rename file

src/llamafactory/data/processor/pretrain.py

@@ -0,0 +1,57 @@
+# Copyright 2024 HuggingFace Inc. and the LlamaFactory team.
+#
+# This code is inspired by the HuggingFace's transformers library.
+# https://github.com/huggingface/transformers/blob/v4.40.0/examples/pytorch/language-modeling/run_clm.py
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from dataclasses import dataclass
+from itertools import chain
+from typing import Any, Dict, List
+
+from .processor_utils import DatasetProcessor
+
+
+@dataclass
+class PretrainDatasetProcessor(DatasetProcessor):
+    def preprocess_dataset(self, examples: Dict[str, List[Any]]) -> Dict[str, List[Any]]:
+        # build grouped texts with format `X1 X2 X3 ...` if packing is enabled
+        eos_token = "<|end_of_text|>" if self.data_args.template == "llama3" else self.tokenizer.eos_token
+        text_examples = [messages[0]["content"] + eos_token for messages in examples["_prompt"]]
+
+        if not self.data_args.packing:
+            if getattr(self.tokenizer, "add_bos_token", False):
+                text_examples = [self.tokenizer.bos_token + example for example in text_examples]
+
+            result = self.tokenizer(
+                text_examples, add_special_tokens=False, truncation=True, max_length=self.data_args.cutoff_len
+            )
+        else:
+            tokenized_examples = self.tokenizer(text_examples, add_special_tokens=False)
+            concatenated_examples = {k: list(chain(*tokenized_examples[k])) for k in tokenized_examples.keys()}
+            total_length = len(concatenated_examples[list(concatenated_examples.keys())[0]])
+            block_size = self.data_args.cutoff_len
+            total_length = (total_length // block_size) * block_size
+            result = {
+                k: [t[i : i + block_size] for i in range(0, total_length, block_size)]
+                for k, t in concatenated_examples.items()
+            }
+            if getattr(self.tokenizer, "add_bos_token", False):
+                for i in range(len(result["input_ids"])):
+                    result["input_ids"][i][0] = self.tokenizer.bos_token_id
+
+        return result
+
+    def print_data_example(self, example: Dict[str, List[int]]) -> None:
+        print("input_ids:\n{}".format(example["input_ids"]))
+        print("inputs:\n{}".format(self.tokenizer.decode(example["input_ids"], skip_special_tokens=False)))