Commit
•
553c80f
1
Parent(s):
eb4c994
streaming multipack for pretraining dataset (#959)
Browse files* [Feat] streaming multipack
* WIP make continued pretraining work w multipack
* fix up hadrcoding, lint
* fix dict check
* update test for updated pretraining multipack code
* fix hardcoded data collator fix for multipack pretraining
* fix the collator to be the max length for multipack pretraining
* don't bother with latest tag for test
* cleanup docker build/test
---------
Co-authored-by: jinwonkim93@github.com <jinwonkim>
Co-authored-by: Wing Lian <wing.lian@gmail.com>
- .github/workflows/tests-docker.yml +2 -4
- examples/tiny-llama/pretrain.yml +58 -0
- src/axolotl/core/trainer_builder.py +14 -4
- src/axolotl/utils/collators.py +21 -0
- src/axolotl/utils/data.py +95 -4
- src/axolotl/utils/trainer.py +10 -0
- tests/test_packed_pretraining.py +82 -0
.github/workflows/tests-docker.yml
CHANGED
@@ -20,7 +20,6 @@ jobs:
|
|
20 |
python_version: "3.10"
|
21 |
pytorch: 2.0.1
|
22 |
axolotl_extras:
|
23 |
-
is_latest: true
|
24 |
- cuda: 121
|
25 |
cuda_version: 12.1.0
|
26 |
python_version: "3.10"
|
@@ -37,7 +36,7 @@ jobs:
|
|
37 |
images: winglian/axolotl
|
38 |
- name: Set up Docker Buildx
|
39 |
uses: docker/setup-buildx-action@v3
|
40 |
-
- name: Build
|
41 |
uses: docker/build-push-action@v5
|
42 |
with:
|
43 |
context: .
|
@@ -49,8 +48,7 @@ jobs:
|
|
49 |
file: ./docker/Dockerfile
|
50 |
tags: |
|
51 |
${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
|
52 |
-
${{ (matrix.is_latest) && format('{0}-latest', steps.metadata.outputs.tags) || '' }}
|
53 |
labels: ${{ steps.metadata.outputs.labels }}
|
54 |
-
- name: Unit Tests
|
55 |
run: |
|
56 |
docker run --rm ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }} pytest --ignore=tests/e2e/ /workspace/axolotl/tests/
|
|
|
20 |
python_version: "3.10"
|
21 |
pytorch: 2.0.1
|
22 |
axolotl_extras:
|
|
|
23 |
- cuda: 121
|
24 |
cuda_version: 12.1.0
|
25 |
python_version: "3.10"
|
|
|
36 |
images: winglian/axolotl
|
37 |
- name: Set up Docker Buildx
|
38 |
uses: docker/setup-buildx-action@v3
|
39 |
+
- name: Build Docker image
|
40 |
uses: docker/build-push-action@v5
|
41 |
with:
|
42 |
context: .
|
|
|
48 |
file: ./docker/Dockerfile
|
49 |
tags: |
|
50 |
${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
|
|
|
51 |
labels: ${{ steps.metadata.outputs.labels }}
|
52 |
+
- name: Unit Tests w docker image
|
53 |
run: |
|
54 |
docker run --rm ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }} pytest --ignore=tests/e2e/ /workspace/axolotl/tests/
|
examples/tiny-llama/pretrain.yml
ADDED
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
|
2 |
+
|
3 |
+
model_type: LlamaForCausalLM
|
4 |
+
tokenizer_type: LlamaTokenizer
|
5 |
+
is_llama_derived_model: true
|
6 |
+
|
7 |
+
load_in_8bit: false
|
8 |
+
load_in_4bit: false
|
9 |
+
strict: false
|
10 |
+
|
11 |
+
max_steps: 200
|
12 |
+
pretraining_dataset:
|
13 |
+
path: c4
|
14 |
+
name: en
|
15 |
+
dataset_prepared_path:
|
16 |
+
val_set_size: 0.0
|
17 |
+
output_dir: ./model-out
|
18 |
+
|
19 |
+
sequence_len: 2048
|
20 |
+
sample_packing: true
|
21 |
+
|
22 |
+
wandb_project:
|
23 |
+
wandb_entity:
|
24 |
+
wandb_watch:
|
25 |
+
wandb_name:
|
26 |
+
wandb_log_model:
|
27 |
+
|
28 |
+
gradient_accumulation_steps: 4
|
29 |
+
micro_batch_size: 2
|
30 |
+
num_epochs: 4
|
31 |
+
optimizer: adamw_bnb_8bit
|
32 |
+
lr_scheduler: cosine
|
33 |
+
learning_rate: 0.0002
|
34 |
+
|
35 |
+
train_on_inputs: false
|
36 |
+
group_by_length: false
|
37 |
+
bf16: true
|
38 |
+
fp16: false
|
39 |
+
tf32: false
|
40 |
+
|
41 |
+
gradient_checkpointing: true
|
42 |
+
early_stopping_patience:
|
43 |
+
resume_from_checkpoint:
|
44 |
+
local_rank:
|
45 |
+
logging_steps: 1
|
46 |
+
xformers_attention:
|
47 |
+
flash_attention: true
|
48 |
+
|
49 |
+
warmup_steps: 10
|
50 |
+
evals_per_epoch:
|
51 |
+
eval_table_size:
|
52 |
+
saves_per_epoch: 1
|
53 |
+
debug:
|
54 |
+
deepspeed:
|
55 |
+
weight_decay: 0.0
|
56 |
+
fsdp:
|
57 |
+
fsdp_config:
|
58 |
+
special_tokens:
|
src/axolotl/core/trainer_builder.py
CHANGED
@@ -60,6 +60,12 @@ class AxolotlTrainingArguments(TrainingArguments):
|
|
60 |
default=False,
|
61 |
metadata={"help": "Use quadratic warmup for cosine scheduling."},
|
62 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
sample_packing: bool = field(
|
64 |
default=False,
|
65 |
metadata={"help": "Use sample packing for efficient training."},
|
@@ -157,7 +163,7 @@ class AxolotlTrainer(Trainer):
|
|
157 |
return self.lr_scheduler
|
158 |
|
159 |
def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
|
160 |
-
if self.args.sample_packing:
|
161 |
return MultipackBatchSampler(
|
162 |
RandomSampler(self.train_dataset),
|
163 |
self.args.train_batch_size,
|
@@ -193,7 +199,7 @@ class AxolotlTrainer(Trainer):
|
|
193 |
return super()._get_eval_sampler(eval_dataset)
|
194 |
|
195 |
def get_train_dataloader(self) -> DataLoader:
|
196 |
-
if self.args.sample_packing:
|
197 |
train_dataset = self.train_dataset
|
198 |
train_dataset = train_dataset.remove_columns(["length"])
|
199 |
data_collator = self.data_collator
|
@@ -768,6 +774,7 @@ class HFCausalTrainerBuilder(TrainerBuilderBase):
|
|
768 |
training_arguments_kwargs
|
769 |
)
|
770 |
training_arguments_kwargs["model_type"] = self.cfg.model_config_type
|
|
|
771 |
|
772 |
if self.cfg.neftune_noise_alpha is not None:
|
773 |
training_arguments_kwargs[
|
@@ -808,7 +815,7 @@ class HFCausalTrainerBuilder(TrainerBuilderBase):
|
|
808 |
train_dataset=self.train_dataset,
|
809 |
eval_dataset=self.eval_dataset,
|
810 |
args=training_args,
|
811 |
-
data_collator=self.build_collator(**data_collator_kwargs),
|
812 |
bench_data_collator=transformers.DataCollatorForSeq2Seq(
|
813 |
self.tokenizer,
|
814 |
return_tensors="pt",
|
@@ -829,7 +836,10 @@ class HFCausalTrainerBuilder(TrainerBuilderBase):
|
|
829 |
|
830 |
return trainer
|
831 |
|
832 |
-
def build_collator(self, **kwargs):
|
|
|
|
|
|
|
833 |
if self.cfg.model_config_type == "mamba":
|
834 |
return MambaDataCollator(tokenizer=self.tokenizer)
|
835 |
|
|
|
60 |
default=False,
|
61 |
metadata={"help": "Use quadratic warmup for cosine scheduling."},
|
62 |
)
|
63 |
+
pretraining: bool = field(
|
64 |
+
default=False,
|
65 |
+
metadata={
|
66 |
+
"help": "Indicates to trainer whether we are doing continued pretraining."
|
67 |
+
},
|
68 |
+
)
|
69 |
sample_packing: bool = field(
|
70 |
default=False,
|
71 |
metadata={"help": "Use sample packing for efficient training."},
|
|
|
163 |
return self.lr_scheduler
|
164 |
|
165 |
def _get_train_sampler(self) -> Optional[torch.utils.data.Sampler]:
|
166 |
+
if self.args.sample_packing and not self.args.pretraining:
|
167 |
return MultipackBatchSampler(
|
168 |
RandomSampler(self.train_dataset),
|
169 |
self.args.train_batch_size,
|
|
|
199 |
return super()._get_eval_sampler(eval_dataset)
|
200 |
|
201 |
def get_train_dataloader(self) -> DataLoader:
|
202 |
+
if self.args.sample_packing and not self.args.pretraining:
|
203 |
train_dataset = self.train_dataset
|
204 |
train_dataset = train_dataset.remove_columns(["length"])
|
205 |
data_collator = self.data_collator
|
|
|
774 |
training_arguments_kwargs
|
775 |
)
|
776 |
training_arguments_kwargs["model_type"] = self.cfg.model_config_type
|
777 |
+
training_arguments_kwargs["pretraining"] = bool(self.cfg.pretraining_dataset)
|
778 |
|
779 |
if self.cfg.neftune_noise_alpha is not None:
|
780 |
training_arguments_kwargs[
|
|
|
815 |
train_dataset=self.train_dataset,
|
816 |
eval_dataset=self.eval_dataset,
|
817 |
args=training_args,
|
818 |
+
data_collator=self.build_collator(training_args, **data_collator_kwargs),
|
819 |
bench_data_collator=transformers.DataCollatorForSeq2Seq(
|
820 |
self.tokenizer,
|
821 |
return_tensors="pt",
|
|
|
836 |
|
837 |
return trainer
|
838 |
|
839 |
+
def build_collator(self, training_args: AxolotlTrainingArguments, **kwargs):
|
840 |
+
if training_args.pretraining:
|
841 |
+
return None
|
842 |
+
|
843 |
if self.cfg.model_config_type == "mamba":
|
844 |
return MambaDataCollator(tokenizer=self.tokenizer)
|
845 |
|
src/axolotl/utils/collators.py
CHANGED
@@ -178,3 +178,24 @@ class MambaDataCollator:
|
|
178 |
"input_ids": input_ids,
|
179 |
"labels": labels,
|
180 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
"input_ids": input_ids,
|
179 |
"labels": labels,
|
180 |
}
|
181 |
+
|
182 |
+
|
183 |
+
@dataclass
|
184 |
+
class PretrainingBatchSamplerDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
|
185 |
+
"""
|
186 |
+
Collator for multipack specific to the using the BatchSampler
|
187 |
+
"""
|
188 |
+
|
189 |
+
def __call__(self, features, return_tensors=None):
|
190 |
+
chunked_data = {}
|
191 |
+
for feature in features.keys():
|
192 |
+
if feature == "length":
|
193 |
+
continue
|
194 |
+
if feature == "attention_mask":
|
195 |
+
arrays = [(1) * np.array(item) for item in features[feature]]
|
196 |
+
chunked_data[feature] = np.concatenate(arrays)
|
197 |
+
else:
|
198 |
+
arrays = [np.array(item) for item in features[feature]]
|
199 |
+
chunked_data[feature] = np.concatenate(arrays)
|
200 |
+
features = [chunked_data]
|
201 |
+
return super().__call__(features, return_tensors=return_tensors)
|
src/axolotl/utils/data.py
CHANGED
@@ -2,6 +2,7 @@
|
|
2 |
import functools
|
3 |
import hashlib
|
4 |
import logging
|
|
|
5 |
from pathlib import Path
|
6 |
from typing import Dict, List, Tuple, Union
|
7 |
|
@@ -14,6 +15,7 @@ from datasets import (
|
|
14 |
load_from_disk,
|
15 |
)
|
16 |
from huggingface_hub import hf_hub_download
|
|
|
17 |
from transformers import PreTrainedTokenizerBase
|
18 |
|
19 |
from axolotl.common.const import DEFAULT_DATASET_PREPARED_PATH
|
@@ -39,11 +41,14 @@ from axolotl.prompters import (
|
|
39 |
SummarizeTLDRPrompter,
|
40 |
UnsupportedPrompter,
|
41 |
)
|
|
|
42 |
from axolotl.utils.dict import DictDefault
|
43 |
from axolotl.utils.distributed import is_main_process, zero_first
|
|
|
44 |
from axolotl.utils.trainer import (
|
45 |
calculate_total_num_steps,
|
46 |
process_datasets_for_packing,
|
|
|
47 |
)
|
48 |
|
49 |
LOG = logging.getLogger("axolotl")
|
@@ -64,9 +69,17 @@ def prepare_dataset(cfg, tokenizer):
|
|
64 |
tokenizer, cfg, DEFAULT_DATASET_PREPARED_PATH
|
65 |
)
|
66 |
else:
|
|
|
|
|
|
|
|
|
|
|
|
|
67 |
train_dataset = load_pretraining_dataset(
|
68 |
-
|
69 |
tokenizer,
|
|
|
|
|
70 |
max_tokens=cfg.sequence_len,
|
71 |
seed=cfg.seed or 42,
|
72 |
)
|
@@ -806,9 +819,27 @@ def encode_pretraining(
|
|
806 |
return ret
|
807 |
|
808 |
|
809 |
-
def load_pretraining_dataset(path, tokenizer, max_tokens=2048, seed=42):
|
810 |
-
|
811 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
812 |
dataset = dataset.shuffle(seed=seed, buffer_size=10_000)
|
813 |
dataset = dataset.map(
|
814 |
encode,
|
@@ -819,3 +850,63 @@ def load_pretraining_dataset(path, tokenizer, max_tokens=2048, seed=42):
|
|
819 |
remove_columns=dataset.features.keys(),
|
820 |
)
|
821 |
return dataset
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
import functools
|
3 |
import hashlib
|
4 |
import logging
|
5 |
+
from collections import defaultdict
|
6 |
from pathlib import Path
|
7 |
from typing import Dict, List, Tuple, Union
|
8 |
|
|
|
15 |
load_from_disk,
|
16 |
)
|
17 |
from huggingface_hub import hf_hub_download
|
18 |
+
from torch.utils.data import RandomSampler
|
19 |
from transformers import PreTrainedTokenizerBase
|
20 |
|
21 |
from axolotl.common.const import DEFAULT_DATASET_PREPARED_PATH
|
|
|
41 |
SummarizeTLDRPrompter,
|
42 |
UnsupportedPrompter,
|
43 |
)
|
44 |
+
from axolotl.utils.collators import PretrainingBatchSamplerDataCollatorForSeq2Seq
|
45 |
from axolotl.utils.dict import DictDefault
|
46 |
from axolotl.utils.distributed import is_main_process, zero_first
|
47 |
+
from axolotl.utils.samplers.multipack import MultipackBatchSampler
|
48 |
from axolotl.utils.trainer import (
|
49 |
calculate_total_num_steps,
|
50 |
process_datasets_for_packing,
|
51 |
+
process_pretraining_datasets_for_packing,
|
52 |
)
|
53 |
|
54 |
LOG = logging.getLogger("axolotl")
|
|
|
69 |
tokenizer, cfg, DEFAULT_DATASET_PREPARED_PATH
|
70 |
)
|
71 |
else:
|
72 |
+
path = cfg.pretraining_dataset
|
73 |
+
name = None
|
74 |
+
if isinstance(cfg.pretraining_dataset, dict):
|
75 |
+
path = cfg.pretraining_dataset["path"]
|
76 |
+
name = cfg.pretraining_dataset["name"]
|
77 |
+
|
78 |
train_dataset = load_pretraining_dataset(
|
79 |
+
path,
|
80 |
tokenizer,
|
81 |
+
cfg,
|
82 |
+
name=name,
|
83 |
max_tokens=cfg.sequence_len,
|
84 |
seed=cfg.seed or 42,
|
85 |
)
|
|
|
819 |
return ret
|
820 |
|
821 |
|
822 |
+
def load_pretraining_dataset(path, tokenizer, cfg, name=None, max_tokens=2048, seed=42):
|
823 |
+
if cfg.sample_packing:
|
824 |
+
collate_fn = PretrainingBatchSamplerDataCollatorForSeq2Seq(
|
825 |
+
tokenizer,
|
826 |
+
return_tensors="pt",
|
827 |
+
padding=True,
|
828 |
+
pad_to_multiple_of=max_tokens * cfg.micro_batch_size,
|
829 |
+
)
|
830 |
+
encode = functools.partial(
|
831 |
+
encode_packed_pretraining,
|
832 |
+
tokenizer,
|
833 |
+
collate_fn,
|
834 |
+
max_seq_length=max_tokens,
|
835 |
+
batch_size=cfg.micro_batch_size,
|
836 |
+
)
|
837 |
+
# set this to 1 so downstream data_loader doesn't try to increase the batch again
|
838 |
+
cfg.micro_batch_size = 1
|
839 |
+
else:
|
840 |
+
encode = functools.partial(encode_pretraining, tokenizer, max_tokens)
|
841 |
+
|
842 |
+
dataset = load_dataset(path, streaming=True, split="train", name=name)
|
843 |
dataset = dataset.shuffle(seed=seed, buffer_size=10_000)
|
844 |
dataset = dataset.map(
|
845 |
encode,
|
|
|
850 |
remove_columns=dataset.features.keys(),
|
851 |
)
|
852 |
return dataset
|
853 |
+
|
854 |
+
|
855 |
+
def encode_packed_pretraining(
|
856 |
+
tokenizer: PreTrainedTokenizerBase,
|
857 |
+
collate_fn,
|
858 |
+
examples: List[str],
|
859 |
+
max_seq_length: int = 2048,
|
860 |
+
batch_size: int = 4,
|
861 |
+
) -> Dict[str, List]:
|
862 |
+
# pylint: disable=duplicate-code
|
863 |
+
# tokenize all the examples
|
864 |
+
# rows get split with stride (overlap)
|
865 |
+
res = tokenizer(
|
866 |
+
examples,
|
867 |
+
truncation=True,
|
868 |
+
max_length=max_seq_length - 1,
|
869 |
+
add_special_tokens=True,
|
870 |
+
return_overflowing_tokens=True,
|
871 |
+
stride=256,
|
872 |
+
)
|
873 |
+
|
874 |
+
input_ids = [seq + [tokenizer.eos_token_id] for seq in res["input_ids"]]
|
875 |
+
attention_mask = [seq + [1] for seq in res["attention_mask"]]
|
876 |
+
|
877 |
+
tokenized_examples = {
|
878 |
+
"input_ids": input_ids,
|
879 |
+
"attention_mask": attention_mask,
|
880 |
+
}
|
881 |
+
|
882 |
+
train_dataset = Dataset.from_dict(tokenized_examples)
|
883 |
+
train_dataset = process_pretraining_datasets_for_packing(
|
884 |
+
train_dataset, max_seq_length
|
885 |
+
)
|
886 |
+
|
887 |
+
sampler = MultipackBatchSampler(
|
888 |
+
RandomSampler(train_dataset),
|
889 |
+
batch_size=batch_size,
|
890 |
+
drop_last=True,
|
891 |
+
batch_max_len=batch_size * max_seq_length,
|
892 |
+
lengths=(
|
893 |
+
train_dataset.data.column("position_ids")
|
894 |
+
.to_pandas()
|
895 |
+
.apply(lambda x: x[-1] + 1)
|
896 |
+
.values
|
897 |
+
),
|
898 |
+
)
|
899 |
+
|
900 |
+
chunked_data = defaultdict(list)
|
901 |
+
|
902 |
+
for data in sampler:
|
903 |
+
features = train_dataset[data]
|
904 |
+
features["labels"] = features["input_ids"].copy()
|
905 |
+
collated_features = collate_fn(features)
|
906 |
+
|
907 |
+
for feature in features.keys():
|
908 |
+
if feature == "length":
|
909 |
+
continue
|
910 |
+
chunked_data[feature].append(collated_features[feature].squeeze(0))
|
911 |
+
|
912 |
+
return chunked_data
|
src/axolotl/utils/trainer.py
CHANGED
@@ -143,6 +143,16 @@ def process_datasets_for_packing(cfg, train_dataset, eval_dataset, tokenizer):
|
|
143 |
return train_dataset, eval_dataset
|
144 |
|
145 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
146 |
def calculate_total_num_steps(cfg, train_dataset, update=True):
|
147 |
if not cfg.total_num_tokens:
|
148 |
total_num_tokens = np.sum(
|
|
|
143 |
return train_dataset, eval_dataset
|
144 |
|
145 |
|
146 |
+
def process_pretraining_datasets_for_packing(train_dataset, sequence_len):
|
147 |
+
drop_long = partial(drop_long_seq, sequence_len=sequence_len)
|
148 |
+
|
149 |
+
train_dataset = train_dataset.filter(drop_long)
|
150 |
+
train_dataset = train_dataset.map(
|
151 |
+
add_position_ids,
|
152 |
+
)
|
153 |
+
return train_dataset
|
154 |
+
|
155 |
+
|
156 |
def calculate_total_num_steps(cfg, train_dataset, update=True):
|
157 |
if not cfg.total_num_tokens:
|
158 |
total_num_tokens = np.sum(
|
tests/test_packed_pretraining.py
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""Module for testing streaming dataset sequence packing"""
|
2 |
+
import unittest
|
3 |
+
from functools import partial
|
4 |
+
|
5 |
+
import torch
|
6 |
+
from datasets import load_dataset
|
7 |
+
from torch.utils.data import DataLoader
|
8 |
+
from transformers import AutoTokenizer
|
9 |
+
|
10 |
+
from axolotl.utils.collators import PretrainingBatchSamplerDataCollatorForSeq2Seq
|
11 |
+
from axolotl.utils.data import encode_packed_pretraining
|
12 |
+
|
13 |
+
|
14 |
+
class TestPacking(unittest.TestCase):
|
15 |
+
"""
|
16 |
+
Test class for packing streaming dataset sequences
|
17 |
+
"""
|
18 |
+
|
19 |
+
def setUp(self) -> None:
|
20 |
+
# pylint: disable=duplicate-code
|
21 |
+
self.tokenizer = AutoTokenizer.from_pretrained("huggyllama/llama-7b")
|
22 |
+
self.tokenizer.pad_token = "</s>"
|
23 |
+
self.max_seq_length = 2048
|
24 |
+
self.batch_size = 2
|
25 |
+
|
26 |
+
def test_packing_stream_dataset(self):
|
27 |
+
# pylint: disable=duplicate-code
|
28 |
+
dataset = load_dataset(
|
29 |
+
"c4",
|
30 |
+
"en",
|
31 |
+
streaming=True,
|
32 |
+
)["train"]
|
33 |
+
|
34 |
+
collate_fn = PretrainingBatchSamplerDataCollatorForSeq2Seq(
|
35 |
+
self.tokenizer,
|
36 |
+
return_tensors="pt",
|
37 |
+
padding=True,
|
38 |
+
pad_to_multiple_of=self.max_seq_length,
|
39 |
+
)
|
40 |
+
|
41 |
+
encode = partial(
|
42 |
+
encode_packed_pretraining,
|
43 |
+
self.tokenizer,
|
44 |
+
collate_fn,
|
45 |
+
max_seq_length=self.max_seq_length,
|
46 |
+
batch_size=self.batch_size,
|
47 |
+
)
|
48 |
+
|
49 |
+
dataset = dataset.map(
|
50 |
+
encode,
|
51 |
+
batched=True,
|
52 |
+
input_columns="text",
|
53 |
+
remove_columns=dataset.features.keys(),
|
54 |
+
)
|
55 |
+
|
56 |
+
trainer_loader = DataLoader(
|
57 |
+
dataset,
|
58 |
+
batch_size=1,
|
59 |
+
collate_fn=None,
|
60 |
+
drop_last=True,
|
61 |
+
)
|
62 |
+
idx = 0
|
63 |
+
for data in trainer_loader:
|
64 |
+
if idx > 10:
|
65 |
+
break
|
66 |
+
assert data["input_ids"].shape == torch.Size(
|
67 |
+
[1, self.batch_size * self.max_seq_length]
|
68 |
+
)
|
69 |
+
assert data["position_ids"].shape == torch.Size(
|
70 |
+
[1, self.batch_size * self.max_seq_length]
|
71 |
+
)
|
72 |
+
assert data["labels"].shape == torch.Size(
|
73 |
+
[1, self.batch_size * self.max_seq_length]
|
74 |
+
)
|
75 |
+
assert data["attention_mask"].shape == torch.Size(
|
76 |
+
[1, self.batch_size * self.max_seq_length]
|
77 |
+
)
|
78 |
+
idx += 1
|
79 |
+
|
80 |
+
|
81 |
+
if __name__ == "__main__":
|
82 |
+
unittest.main()
|