WIP

Signed-off-by: Kyle Sayers <[email protected]>

Author: kylesayrs

src/llmcompressor/transformers/finetune/data/data_args.py

@@ -1,8 +1,6 @@
 from dataclasses import dataclass, field
 from typing import Any, Callable, Dict, List, Optional, Union
 
-from transformers import DefaultDataCollator
-
 
 @dataclass
 class DVCDatasetTrainingArguments:
@@ -60,9 +58,12 @@ class CustomDataTrainingArguments(DVCDatasetTrainingArguments):
         },
     )
 
-    data_collator: Callable[[Any], Any] = field(
-        default_factory=lambda: DefaultDataCollator(),
-        metadata={"help": "The function to used to form a batch from the dataset"},
+    data_collator: Optional[Callable[[Any], Any]] = field(
+        default=None,
+        metadata={
+            "help": "The function to used to form a batch from the dataset. Defaults "
+            "to `DataCollatorWithPadding` with model tokenizer if None is provided"
+        },
     )
 
 

src/llmcompressor/transformers/finetune/data/data_helpers.py

@@ -1,11 +1,17 @@
 import logging
 import os
+import warnings
 from typing import Any, Callable, Dict, List, Optional
 
 import torch
 from datasets import Dataset, load_dataset
 from torch.utils.data import DataLoader, RandomSampler, SequentialSampler
-from transformers.data import default_data_collator
+from transformers.data.data_collator import (
+    DataCollatorWithPadding,
+    default_data_collator,
+)
+
+from llmcompressor.typing import Processor
 
 LOGGER = logging.getLogger(__name__)
 LABELS_MASK_VALUE = -100
@@ -21,7 +27,9 @@
 def format_calibration_data(
     tokenized_dataset: Dataset,
     num_calibration_samples: Optional[int] = None,
+    batch_size: int = 1,
     do_shuffle: bool = True,
+    processor: Optional[Processor] = None,
     collate_fn: Callable = default_data_collator,
     accelerator: Optional[Any] = None,
 ) -> List[torch.Tensor]:
@@ -37,6 +45,11 @@ def format_calibration_data(
     :param accelerator: optional accelerator for if preparing in FSDP mode
     :return: list of trimmed calibration data tensors
     """
+    # shuffle
+    if do_shuffle:
+        tokenized_dataset = tokenized_dataset.shuffle()
+
+    # truncate samples
     safe_calibration_samples = len(tokenized_dataset)
     if num_calibration_samples is not None:
         safe_calibration_samples = min(len(tokenized_dataset), num_calibration_samples)
@@ -45,13 +58,22 @@ def format_calibration_data(
                 f"Requested {num_calibration_samples} calibration samples but "
                 f"the provided dataset only has {safe_calibration_samples}. "
             )
-
-    if do_shuffle:
-        tokenized_dataset = tokenized_dataset.shuffle()
     tokenized_calibration = tokenized_dataset.select(range(safe_calibration_samples))
 
+    # collate data
+    if collate_fn is None:
+        tokenizer = getattr(processor, "tokenizer", processor)
+        if tokenizer is None:
+            warnings.warn(
+                "Could not find processor, attempting to collate with without padding "
+                "(may fail for batch_size > 1)"
+            )
+            return default_data_collator()
+
+        collate_fn = DataCollatorWithPadding(tokenizer)
+
     dataloader_params = {
-        "batch_size": 1,
+        "batch_size": batch_size,
         "sampler": RandomSampler(tokenized_calibration)
         if do_shuffle
         else SequentialSampler(tokenized_calibration),

src/llmcompressor/transformers/finetune/runner.py

@@ -144,8 +144,10 @@ def one_shot(self, stage: Optional[str] = None):
             calib_data = format_calibration_data(
                 tokenized_dataset=self.get_dataset_split("calibration"),
                 num_calibration_samples=self._data_args.num_calibration_samples,
+                batch_size=self._training_args.per_device_oneshot_batch_size,
                 do_shuffle=self._data_args.shuffle_calibration_samples,
                 collate_fn=self._data_args.data_collator,
+                processor=self.processor,
                 accelerator=self.trainer.accelerator,
             )
 

src/llmcompressor/transformers/finetune/training_args.py

@@ -32,6 +32,12 @@ class TrainingArguments(HFTrainingArgs):
             )
         },
     )
+    per_device_oneshot_batch_size: int = field(
+        default=1,
+        metadata={
+            "help": "The batch size per GPU/XPU/TPU/MPS/NPU core/CPU for oneshot"
+        },
+    )
     save_compressed: Optional[bool] = field(
         default=True,
         metadata={"help": "Whether to compress sparse models during save"},

tests/llmcompressor/transformers/finetune/data/test_dataset_helpers.py

@@ -1,7 +1,11 @@
+# TODO: rename to `test_data_helpers.py`
 import pytest
+import torch
+from datasets import Dataset
 
 from llmcompressor.transformers.finetune.data.data_args import DataTrainingArguments
 from llmcompressor.transformers.finetune.data.data_helpers import (
+    format_calibration_data,
     get_raw_dataset,
     make_dataset_splits,
 )
@@ -53,3 +57,18 @@ def test_separate_datasets():
         split_datasets = make_dataset_splits(
             datasets, do_train=True, do_eval=True, do_predict=True
         )
+
+
+@pytest.mark.unit
+def test_format_calibration_data():
+    tokenized_dataset = Dataset.from_dict(
+        {"input_ids": torch.randint(0, 512, (8, 2048))}
+    )
+
+    calibration_dataloader = format_calibration_data(
+        tokenized_dataset, num_calibration_samples=4, batch_size=2
+    )
+
+    batch = next(iter(calibration_dataloader))
+
+    assert batch["input_ids"].size(0) == 2