Normal padding (no packing) in SFT

Author: joecummings

apps/sft/llama3_8b.yaml

@@ -27,7 +27,7 @@ lr_scheduler:
   warmup_steps: 200
 
 training:
-  local_batch_size: 1
+  local_batch_size: 8
   seq_len: 2048
   max_norm: 1.0
   steps: 1000

apps/sft/main.py

@@ -16,15 +16,13 @@
 import math
 import os
 import sys
-from functools import partial
 from typing import Any
 
 import torch
 
 import torchtitan.experiments.forge.train_spec as forge_train_spec
 from forge.controller import ForgeActor
-from forge.data.collate import collate_packed
-from forge.data.datasets.packed import PackedDataset, TextPacker
+from forge.data.collate import collate_padded
 from forge.data.datasets.sft_dataset import AlpacaToMessages, sft_iterable_dataset
 from forge.data.tokenizer import HuggingFaceModelTokenizer
 from forge.data.utils import StopAfterOneEpoch
@@ -197,25 +195,12 @@ def setup_data(self, dataset_configs: list[dict]) -> StatefulDataLoader:
             **dataset_config,
         )
 
-        packer = TextPacker(padding_idx=0)
-        dataset = PackedDataset(
-            dataset=dataset,
-            packer=packer,
-            target_tokens_per_pack=self.job_config.training.seq_len,  # TODO: get this from model
-        )
-
         dataloader = StatefulDataLoader(
             dataset=dataset,
             batch_size=self.job_config.training.local_batch_size,
-            collate_fn=partial(
-                collate_packed, mask_fn=packer.create_block_mask, device=self.device
-            ),
+            collate_fn=collate_padded,
         )
 
-        # Ultimately we probably want something like this
-        # packer = build_packing_strategy(packing_config)
-        # dataset = build_dataset(dataset_config)
-        # dataloader = build_dataloader(dataloader_config, dataset, packer)
         return dataloader
 
     def forward_backward(

apps/sft/qwen3_8b.yaml

@@ -26,7 +26,7 @@ lr_scheduler:
   warmup_steps: 200
 
 training:
-  local_batch_size: 1
+  local_batch_size: 8
   seq_len: 2048
   max_norm: 1.0
   steps: 1000

src/forge/data/__init__.py

@@ -3,12 +3,13 @@
 #
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
-from .collate import collate_packed
+from .collate import collate_packed, collate_padded
 from .metric_transform import DefaultDatasetMetricTransform, MetricTransform
 from .utils import CROSS_ENTROPY_IGNORE_IDX
 
 __all__ = [
     "collate_packed",
+    "collate_padded",
     "CROSS_ENTROPY_IGNORE_IDX",
     "MetricTransform",
     "DefaultDatasetMetricTransform",

src/forge/data/collate.py

@@ -7,6 +7,63 @@
 from typing import Any, Callable
 
 import torch
+import torch.nn.functional as F
+
+from forge.data.utils import CROSS_ENTROPY_IGNORE_IDX
+
+
+def collate_padded(batch: list[dict[str, Any]]) -> dict[str, Any]:
+    """
+    Collate function that pads sequences to the longest sample in the batch.
+
+    Pads 'tokens' with 0 and 'labels' with CROSS_ENTROPY_IGNORE_IDX (-100).
+    Non-tensor fields (like metrics) are collected into lists and flattened
+    if all items are lists.
+
+    Args:
+        batch: List of samples, each containing 'tokens' and 'labels' tensors
+
+    Returns:
+        Batched dict with padded tensors
+    """
+    if not batch:
+        return {}
+
+    # Find max length in batch
+    max_len = max(sample["tokens"].size(0) for sample in batch)
+
+    # Initialize lists for batched tensors
+    tokens_list = []
+    labels_list = []
+
+    # Pad each sample to max_len
+    for sample in batch:
+        seq_len = sample["tokens"].size(0)
+        pad_len = max_len - seq_len
+
+        # Pad tokens with 0
+        padded_tokens = F.pad(sample["tokens"], (0, pad_len), value=0)
+        tokens_list.append(padded_tokens)
+
+        # Pad labels with CROSS_ENTROPY_IGNORE_IDX (-100)
+        padded_labels = F.pad(sample["labels"], (0, pad_len), value=CROSS_ENTROPY_IGNORE_IDX)
+        labels_list.append(padded_labels)
+
+    # Stack into batch
+    result = {
+        "tokens": torch.stack(tokens_list),
+        "labels": torch.stack(labels_list),
+    }
+
+    # Collect non-tensor fields (like metrics)
+    for key in batch[0].keys():
+        if key not in ["tokens", "labels"]:
+            result[key] = [sample[key] for sample in batch]
+            # Flatten if all are lists
+            if all(isinstance(item, list) for item in result[key]):
+                result[key] = [item for sublist in result[key] for item in sublist]
+
+    return result
 
 
 def collate_packed(

tests/unit_tests/datasets/test_packed.py

@@ -13,7 +13,8 @@
 import pytest
 import torch
 
-from forge.data.collate import collate_packed
+from forge.data import CROSS_ENTROPY_IGNORE_IDX
+from forge.data.collate import collate_packed, collate_padded
 from forge.data.datasets import HfIterableDataset
 from forge.data.datasets.packed import (
     _SUPPORTS_FLEX_ATTENTION,
@@ -995,3 +996,133 @@ def test_iter_restart_determinism(self, dataset_factory):
                 pack2["document_ids"],
                 msg=f"Pack {i}: document_ids mismatch between iterations",
             )
+
+
+class TestCollatePadded:
+    """Test collate_padded function"""
+
+    def test_empty_batch(self):
+        """Test collating an empty batch"""
+        result = collate_padded([])
+        assert result == {}
+
+    def test_single_sample(self):
+        """Test collating a single sample"""
+        batch = [
+            {
+                "tokens": torch.tensor([1, 2, 3]),
+                "labels": torch.tensor([4, 5, 6]),
+            }
+        ]
+        result = collate_padded(batch)
+
+        assert result["tokens"].shape == (1, 3)
+        assert result["labels"].shape == (1, 3)
+        torch.testing.assert_close(result["tokens"], torch.tensor([[1, 2, 3]]))
+        torch.testing.assert_close(result["labels"], torch.tensor([[4, 5, 6]]))
+
+    def test_equal_length_samples(self):
+        """Test collating samples with equal lengths"""
+        batch = [
+            {
+                "tokens": torch.tensor([1, 2, 3]),
+                "labels": torch.tensor([4, 5, 6]),
+            },
+            {
+                "tokens": torch.tensor([7, 8, 9]),
+                "labels": torch.tensor([10, 11, 12]),
+            },
+        ]
+        result = collate_padded(batch)
+
+        assert result["tokens"].shape == (2, 3)
+        assert result["labels"].shape == (2, 3)
+        torch.testing.assert_close(
+            result["tokens"], torch.tensor([[1, 2, 3], [7, 8, 9]])
+        )
+        torch.testing.assert_close(
+            result["labels"], torch.tensor([[4, 5, 6], [10, 11, 12]])
+        )
+
+    def test_padding_to_longest(self):
+        """Test padding shorter sequences to the longest in batch"""
+        batch = [
+            {
+                "tokens": torch.tensor([1, 2]),
+                "labels": torch.tensor([3, 4]),
+            },
+            {
+                "tokens": torch.tensor([5, 6, 7, 8]),
+                "labels": torch.tensor([9, 10, 11, 12]),
+            },
+            {
+                "tokens": torch.tensor([13, 14, 15]),
+                "labels": torch.tensor([16, 17, 18]),
+            },
+        ]
+        result = collate_padded(batch)
+
+        # All should be padded to length 4 (longest)
+        assert result["tokens"].shape == (3, 4)
+        assert result["labels"].shape == (3, 4)
+
+        # Check tokens padding (padded with 0)
+        torch.testing.assert_close(
+            result["tokens"],
+            torch.tensor([[1, 2, 0, 0], [5, 6, 7, 8], [13, 14, 15, 0]]),
+        )
+
+        # Check labels padding (padded with CROSS_ENTROPY_IGNORE_IDX)
+        torch.testing.assert_close(
+            result["labels"],
+            torch.tensor(
+                [
+                    [3, 4, CROSS_ENTROPY_IGNORE_IDX, CROSS_ENTROPY_IGNORE_IDX],
+                    [9, 10, 11, 12],
+                    [16, 17, 18, CROSS_ENTROPY_IGNORE_IDX],
+                ]
+            ),
+        )
+
+    def test_non_tensor_fields_preserved(self):
+        """Test that non-tensor fields are collected correctly"""
+        batch = [
+            {
+                "tokens": torch.tensor([1, 2]),
+                "labels": torch.tensor([3, 4]),
+                "metadata": "sample1",
+            },
+            {
+                "tokens": torch.tensor([5, 6, 7]),
+                "labels": torch.tensor([8, 9, 10]),
+                "metadata": "sample2",
+            },
+        ]
+        result = collate_padded(batch)
+
+        assert "metadata" in result
+        assert result["metadata"] == ["sample1", "sample2"]
+
+    def test_metrics_flattened(self):
+        """Test that metrics lists are flattened"""
+        batch = [
+            {
+                "tokens": torch.tensor([1, 2]),
+                "labels": torch.tensor([3, 4]),
+                "metrics": [
+                    type("Metric", (), {"key": "loss", "value": 1.0})(),
+                    type("Metric", (), {"key": "acc", "value": 0.9})(),
+                ],
+            },
+            {
+                "tokens": torch.tensor([5, 6, 7]),
+                "labels": torch.tensor([8, 9, 10]),
+                "metrics": [type("Metric", (), {"key": "loss", "value": 2.0})()],
+            },
+        ]
+        result = collate_padded(batch)
+
+        assert "metrics" in result
+        # Should be flattened from [[metric1, metric2], [metric3]] to [metric1, metric2, metric3]
+        assert len(result["metrics"]) == 3
+